premiere-pro-mcp

What is this?

MCP for Adobe Premiere Pro is the independent, MIT-licensed local MCP server published as premiere-pro-mcp by leancoderkavy/premiere-pro-mcp. Install this package and its separate Premiere connector to connect a compatible AI assistant to supported project inspection, timeline editing, and export tools. Start with verify_premiere_connection and make no changes until the bridge works.

Find the right Adobe Premiere Pro MCP guide

An MCP (Model Context Protocol) server that lets AI assistants like Claude, Windsurf, Cursor, GitHub Copilot, or any MCP-compatible client directly control Adobe Premiere Pro — importing media, editing timelines, applying effects, managing keyframes, exporting, and more.

"Add the B-roll clips to V2, apply a cross dissolve between each, color correct them to match the A-roll, and export a 1080p ProRes."

At a glance

This project is independent and is not affiliated with or endorsed by Adobe Inc. It is also separate from other MCP servers for Premiere Pro. The only npm package published from this repository is premiere-pro-mcp, and no other package name installs it.

Comparing similarly named packages? Both this package and adobe-premiere-pro-mcp declare a premiere-pro-mcp executable. Verify the package and repository before configuring a client. The new local configuration helper prints Claude, Cursor, VS Code, or Codex settings that point directly to this installation. It is a feature included in v1.15.1 and later.

The current source exposes 393 core tools for supported workflow steps spanning the supported ExtendScript, QE DOM, local media and interchange analysis, revisioned project-context retrieval, safe edit-planning, project-intake preview, review handoff, connection verification, and guarded After Effects MOGRT authoring, batch, library, render-queue, inspection, and Premiere-handoff workflows. A compatible, authenticated UXP panel adds 96 documented, capability-gated tools without replacing the production CEP bridge.

Latest release: 1.22.0

The published v1.22.0 npm artifact contains 393 core tools, 391 in its default profile, and 487 with a compatible UXP connection. The development catalog above can include unreleased work. See the versioned facts and package provenance.

GitHub releases use v<SemVer> for both title and tag, such as v1.22.0. See release naming and version conventions.

Try a bounded workflow

Choose a setup guide: Claude Desktop, Codex, other local MCP clients, or ChatGPT connection options.

Download the workflow starter kit for synthetic media and three evaluation recipes: a read-only sequence check, an explicitly confirmed review-frame export, and a product-spot preview. No email is required. The kit has no recorded Premiere demo or verified host result. Start with a disposable project and use setup and recovery if the connection is unavailable.

Release highlights

• 1.22.0 readback receipts: effect writes, stabilization, smart bins, proxies, folder imports, sequence creation, marker deletion and clip selection report what Premiere applied as verified, committed_unverified or failed; create_subclip, add_title and update_marker report the stored values; set_sequence_frame_rate lists clips Premiere re-snapped.

• 1.22.0 display codes and proxies: set_sequence_display_format writes Premiere's real display codes (video 100-113, audio 200-201) and maps the older 0-11 and 0-1 inputs onto them; manage_proxies toggle can set a known proxy display state; the Windows ACL check waits up to 30 seconds for PowerShell.

• 1.21.0 Claude Desktop startup fix: the extension starts when optional settings are blank, and Desktop's discover-only probe no longer holds the UXP port, so the real server keeps its UXP tools.

• 1.21.0 frame-grid verification: project-item marks, sequence In/Out, work area, link/unlink, zero point, transitions, replace_clip and duplicate_clip verify against Premiere's own frame and sample grids and fail closed; UXP tools snap times to the sequence frame grid.

• 1.20.0 verified range removal: ripple_remove_timeline_ranges previews, then applies a multi-range ripple removal with a single-use confirmation token and clip-level readback (QE razor; experimental).

• 1.20.0 effect safety: effect colours read and write losslessly, ambiguous property names (for example repeated Lumetri controls) are refused unless property_index is given, and duplicate_clip copies keep the original length.

• 1.20.0 import and metadata additions: import_fcp_xml can import into the open project, import_folder accepts target_bin, get_encoder_presets pages with limit/offset, set_footage_interpretation sets field_type, and XML/AAF/EDL imports through import_media no longer report a false error.

• 1.19.1 timeline fixes: bounded sequence pages with continuation, measured unlinked-clip compatibility, and faster ripple movement with a large-edit opt-in and scaled host waits.

• 1.19.1 safer handoffs: AME queue-wide start requires explicit opt-in, malformed MOGRT files refuse before host import, and uncertain committed edits retain inspection guidance.

• 1.19.1 transcript and capture fixes: empty host-attached transcripts can accept guarded imports; stored values and still files do not establish temporal animation. A short actual video export is needed to assess keyframe curves.

• Editorial planning: film evidence, transcript cleanup, caption authoring, shorts and chapter planning, rhythm and speaker layouts, timeline QA, and platform delivery plans.

• Cross-app handoff: capability-aware workflow routes and guarded After Effects render-to-Premiere handoff.

• Editing correctness: verified readback and explicit capability or committed-but-unverified failures across transition, import, effects, source identity, and track operations. Insert edits ripple QE sync-locked tracks instead of silently desyncing neighbours.

• MOGRT studio: an optional, separate After Effects CEP connector can author approval-gated title, callout, quote, and social recipes; constrain them with local brand kits; batch, publish, queue, inspect, and hand them to Premiere without claiming visual or completed-render proof.

• Global-update handoff: a global npm installation can show its server and connector update state in the Windows CEP panel and, after explicit confirmation, update only after Premiere has been quit; it never changes a project, client configuration, source checkout, or custom npm prefix.

• Local review planning: caption timing previews parse caller-supplied SRT/VTT without writing or importing it, while revision-bound editorial evidence stays in the opt-in local context index without provider calls.

• Repair preview: premiere-pro-mcp --doctor --plan-fixes produces a privacy-safe, no-write repair plan; the narrowly eligible connector repair still requires an explicit confirmation that Premiere is closed.

• Auditable guidance: the universal setup guide, generated public workflow manifest, and proof runbook make workflow and verification boundaries inspectable without claiming a licensed-host walkthrough occurred.

• Faster, clearer delivery: immutable MCP registration work is reused safely across stateless requests, and the website gained a lighter, mobile-first workflow view plus current machine-readable facts and crawl guidance for its public pages.

• Explicit boundary: the hosted endpoint remains an operator-managed MCP service; unauthenticated callers are rejected and it does not pair users to local Premiere processes. See the generated supported action catalog for individual capability and verification contracts.

See the v1.22.0 release notes for complete details. Live installation in Premiere Pro still requires host verification.

Current MCP protocol support

The server uses the stable TypeScript SDK v2 and serves the 2026-07-28 stateless protocol over HTTP and stdio, while retaining legacy MCP compatibility through 2025-11-25. Modern clients receive discovery, validated routing headers, cache hints, subscription-stream support, and the formal Premiere extension capability. See the complete MCP capability and boundary report.

If an MCP client cannot complete its server/discover probe, set PREMIERE_MCP_PROTOCOL_MODE=legacy in that client's server environment and restart the client. This uses the SDK's base stdio transport and the legacy initialization handshake only; leave it unset (or auto) for modern MCP capabilities.

For editors evaluating an AI workflow

Before an assistant changes an active project, use the Premiere Pro AI workflow checklist to define the target and no-change boundaries, verify the local connection, request a bounded plan, and inspect the returned result. It is a practical starting point for assistant editors and post leads testing a repeatable workflow on a duplicate project or small test sequence.

For product context, see the Adobe Premiere AI Assistant and MCP comparison and the Claude Desktop setup guide.

If you are deciding between a single local project, an Adobe Production on shared storage, or a remote Team Project, use the Premiere Pro collaboration workflow guide before you evaluate an MCP path. It links the relevant Adobe guidance, makes no project inspection request, and ends with the same read-only connection check.

For a concrete first Project Intake preview, choose one of the three schema-checked, no-sensitive-data starter templates. They are evaluation samples only: a human policy owner must review and replace their bins, media rules, and organization rules before a facility uses one.

Quick Start

Install the published package (verify the name)

npm i -g [email protected]

This repository publishes only premiere-pro-mcp. A differently named package (adobe-premiere-pro-mcp) may also declare a premiere-pro-mcp executable. Before configuring a client, confirm:

Check	Expected
Package name	premiere-pro-mcp (not adobe-premiere-pro-mcp)
Version	1.22.0
Homepage / repo	https://premiere-pro-mcp.com/ · https://github.com/leancoderkavy/premiere-pro-mcp

npm list -g premiere-pro-mcp
premiere-pro-mcp --version
npm view premiere-pro-mcp homepage repository.url

Then continue with --install-cep, --doctor, and the read-only verify_premiere_connection prompt.

Easiest supported path: Claude Desktop

1. Download the current Claude Desktop bundle (.mcpb).
2. In Claude Desktop, open Settings > Extensions > Advanced settings > Install Extension, select the downloaded bundle, and restart Claude Desktop.
3. Download the separate signed Premiere connector (.zxp). Open it with your trusted ZXP installer. If your computer has no ZXP installer, use the npm connector installer in Advanced setup below.
4. Restart Premiere, open a project, then open Window > Extensions > MCP for Adobe Premiere Pro.
5. In Claude, enter: Safely check my Premiere connection with verify_premiere_connection. Make no changes.

The Claude bundle contains the local MCP server, so this route does not require Node.js. The Premiere connector is a separate required install. The first prompt is read-only and reports whether the server is installed, configured, connected, and live-verified.

First proof, before the first edit

This is an illustrated workflow, not a Premiere panel screenshot or licensed-host proof.

1. Open a copied test project and an active sequence in Premiere.
2. Open Window > Extensions > MCP for Adobe Premiere Pro. “Running” means the local panel bridge is available; it does not show that an edit completed.
3. Run premiere-pro-mcp --doctor to check only local package/configuration readiness.
4. Ask the AI client: Run verify_premiere_connection. Make no changes. The returned check is read-only and avoids project names, paths, and media details.

If a bridge, project, or active sequence is missing, fix that setup state before allowing a mutation. For a concise, translatable version of this path, see quick starts in English, Spanish, and Japanese. The translations are machine-assisted drafts and retain command names in English.

Other AI assistants

Cursor, VS Code/Copilot, Windsurf, and other MCP clients do not currently have a project-provided one-click installer. Use their MCP settings with the advanced npm route below. Keep the assistant, server, connector, and Premiere on the same computer.

For a portable reference users can download and attach to any AI assistant, see the MCP for Adobe Premiere Pro setup guide for AI assistants. Attaching the guide provides assistant context; the local server and Premiere connector still need to be installed separately.

npm install -g premiere-pro-mcp

git clone https://github.com/leancoderkavy/premiere-pro-mcp.git
cd premiere-pro-mcp
npm install
npm run build

premiere-pro-mcp --install-cep

npm run install-cep

premiere-pro-mcp --doctor

premiere-pro-mcp --check-update
premiere-pro-mcp --update

npm run check-update:source
npm run update:source

premiere-pro-mcp --uninstall-cep

Publishing to npm

The easiest repeatable path is the token-free GitHub Actions workflow:

1. In the npm package settings, configure GitHub Actions as the trusted publisher for leancoderkavy/premiere-pro-mcp and workflow file npm-publish.yml.
2. Allow the npm publish action.
3. Open Actions -> Publish npm -> Run workflow and keep the default latest tag.

The workflow installs dependencies, builds, runs tests, verifies the packed files, refuses to republish an existing version, then publishes through short-lived OIDC credentials with automatic provenance. No npm token or recurring OTP is required.

For local publishing, use the guided helper:

npm run publish:npm

Useful local variants:

npm run publish:npm:dry-run
NPM_OTP=123456 npm run publish:npm
NPM_TOKEN=npm_xxx npm run publish:npm

mkdir -p ~/Library/Application\ Support/Adobe/CEP/extensions
ln -s "$(pwd)/cep-plugin" ~/Library/Application\ Support/Adobe/CEP/extensions/MCPBridgeCEP

# Enable unsigned extensions (CSXS 9–14)
for v in 9 10 11 12 13 14; do
  defaults write com.adobe.CSXS.$v PlayerDebugMode 1
done

3. Configure your MCP client

If you installed from npm, configure the client to run the global command:

{
  "mcpServers": {
    "premiere-pro": {
      "command": "premiere-pro-mcp"
    }
  }
}

If you cloned the repository instead, use the source-build configuration shown below for your client.

{
  "mcpServers": {
    "premiere-pro": {
      "command": "node",
      "args": ["/absolute/path/to/premiere-pro-mcp/dist/index.js"]
    }
  }
}

{
  "premiere-pro": {
    "command": "node",
    "args": ["/absolute/path/to/premiere-pro-mcp/dist/index.js"]
  }
}

{
  "mcpServers": {
    "premiere-pro": {
      "command": "node",
      "args": ["/absolute/path/to/premiere-pro-mcp/dist/index.js"]
    }
  }
}

{
  "mcpServers": {
    "premiere-pro": {
      "command": "node",
      "args": ["/absolute/path/to/premiere-pro-mcp/dist/index.js"]
    }
  }
}

4. Verify the bridge in Premiere Pro

1. Open (or restart) Premiere Pro
2. The bridge starts automatically using the default temp directory (or its previously saved setting)
3. Optionally go to Window > Extensions > MCP for Adobe Premiere Pro to confirm the green "Running" status or change the Temp Directory to match your MCP client config
4. Ask your AI assistant to run get_capabilities, then ping, with Premiere open.
5. For a safe first request, ask: "What is my current Premiere Pro project and active sequence? Do not make changes."

The default bridge directory is derived from the operating system on both sides, so most local setups should not set PREMIERE_TEMP_DIR. On macOS, the server resolves the per-user system temporary directory even when a GUI-launched client strips TMPDIR, keeping it aligned with Premiere's CEP panel. If you override it, use the same absolute path in the MCP server and CEP panel; Windows and macOS paths are not interchangeable.

Codex plugin

This repository includes an installable Codex plugin that bundles the local MCP server with a safety-oriented Premiere editing skill.

From a clone of this repository:

codex plugin marketplace add .
codex plugin add premiere-pro@premiere-pro-mcp
npx -y [email protected] --install-cep

Restart Premiere Pro and start a new Codex session after installation. The plugin launches [email protected] through npx; the separate CEP installation is required because the MCP server communicates with the running Premiere host through the local bridge.

The plugin source lives in plugins/premiere-pro, and the repository marketplace manifest lives in .agents/plugins/marketplace.json.

GPT-6 Astra and agent tool discovery

Use the Codex plugin with codex --model gpt-6-astra when your account has access. The server supplies session-aware workflow instructions and bounded tool discovery: call get_capabilities with {"tool_query":"transcript","tool_limit":10} to find relevant operations, their descriptions, and backend requirements. Searches default to tools registered under the current authority and tool packs.

See GPT-6 Astra workflows for evidence retrieval, visual review, execution ordering, and the division between MCP and client capabilities.

GPT-6 Sol and GPT-6 Luna

Use the Codex plugin with codex --model gpt-6-sol for multi-step agentic edits, or codex --model gpt-6-luna for focused, high-volume inspection passes, when your account has access. Both call the same MCP tools under the same authority, preview, and readback rules. See GPT-6 Sol and Luna workflows.

Claude

For Claude Code, add this repository as a marketplace and install the plugin:

/plugin marketplace add leancoderkavy/premiere-pro-mcp
/plugin install premiere-pro@premiere-pro-mcp

Then install the Premiere bridge and start a new Claude Code session:

npx -y [email protected] --install-cep

The Claude Code package lives in claude-plugins/premiere-pro, with its marketplace at .claude-plugin/marketplace.json.

Claude Desktop uses the self-contained MCP Bundle (.mcpb) format. Build and validate the current bundle with:

npm run build:claude

Install the resulting file from artifacts/ through Settings > Extensions > Advanced settings > Install Extension. The Premiere CEP bridge must still be installed separately.

Claude Fable 5.1

Use Cursor, Claude Desktop, or Claude Code with Claude Fable 5.1 (claude-fable-5-1) when your account has access. Model selection belongs to the client; this server does not run an Anthropic model. Fable 5.1 is optional: other Claude models can call the same MCP tools. If Cursor Privacy Mode or an Enterprise plan is enabled, an admin must approve Anthropic's Fable data-retention policy before the model is available.

The server supplies session-aware workflow instructions and bounded tool discovery: call get_capabilities with {"tool_query":"transcript","tool_limit":10} to find relevant operations, their descriptions, and backend requirements. Keep Premiere mutations serialized even in a long Fable 5.1 session. Image review of returned frames is not playback or delivery proof.

See Claude Fable 5.1 workflows for connection order, privacy boundaries, evidence retrieval, Cursor's Opus fallback, and the division between MCP and client capabilities. The public walkthrough is How to Use Claude Fable 5.1 with MCP for Adobe Premiere Pro.

Claude Opus 5.5

Use Claude Code, Claude Desktop, or Cursor with Claude Opus 5.5 (claude-opus-5-5) when your account has access. Model selection belongs to the client; this server does not run an Anthropic model. Opus 5.5 is optional. Keep Premiere mutations serialized and preview before apply. See Claude Opus 5.5 workflows.

Windows and macOS capability coverage

get_capabilities reports the current operating system, temp directory, CEP/UXP coverage, enabled authority profile, and any live-host verification still required. It also includes the full tools catalog generated from the tools registered by the server, including tools disabled by the active profile. Every entry identifies:

• the execution backend (local, CEP/ExtendScript, QE, or orchestrator);
• static support status (supported, limited, experimental, or unsupported);
• the minimum Premiere version known to the server;
• the required authority and whether the current profile enables it;
• the verification boundary and whether a live Premiere host is required; and
• relevant operational notes.

QE-backed tools are reported as experimental because QE is undocumented and can vary between Premiere builds. Authority availability is reported separately from implementation support, so disabling edit, for example, does not incorrectly label editing tools as unsupported. Static metadata never claims that a Premiere operation succeeded; use ping and inspect each tool result for runtime evidence.

MCP tools/list is filtered to the active authority profile. The default inspect,edit,export,filesystem profile advertises 367 of the 369 registered tools and omits execute_extendscript and evaluate_expression, which require explicit unsafe-script authority. ping and get_capabilities remain visible under every profile so a restricted or misconfigured server can still explain its state. The call-time capability guard remains authoritative even if listing metadata is wrong.

Workflow-scoped discovery packs and structured outputs

By default, the full permitted catalog remains available. Set PREMIERE_MCP_TOOL_PACKS to essential, inspection, delivery, captions, or a comma-separated combination such as inspection,captions to reduce the tool discovery and registered session surface for a focused client. full is the explicit full-catalog mode and cannot be combined with another pack. ping and get_capabilities remain listed for diagnosis, and every registered call still passes through the same capability guard; a pack never grants authority.

Every listed tool now declares the same machine-readable result envelope through MCP outputSchema: ok, tool, plus data on success or error on failure. The tool-specific data shape remains versioned by the individual tool result, so clients can reliably distinguish transport success from a Premiere or local operation failure without parsing the text block.

After Effects MOGRT studio

This is a narrow authoring path, not an arbitrary After Effects script runner. Install the separate local connector, fully restart After Effects, and open Window > Extensions > MCP for Adobe After Effects:

premiere-pro-mcp --install-after-effects-cep

Then open a saved .aep project inside an approved workspace and use this order:

1. verify_after_effects_connection — read-only connector and saved-project check.
2. preview_mogrt_recipe — produces an expiring, one-time plan for one of five supported recipes: lower_third, title_card, callout, quote_card, or social_end_card; it never creates directories or contacts Adobe.
3. create_mogrt_recipe with that token and confirm_export: true — creates one composition, saves the open project, and requests one .mogrt export.
4. verify_mogrt_artifact — checks only local file existence and its ZIP header.

Optional studio paths retain the same preview-and-confirm boundary:

• validate_mogrt_brand_kit validates local name-prefix, accent/text colors, font request, safe-margin, and workspace-contained logo values before they are passed into a recipe.
• preview_mogrt_batch / create_mogrt_batch export up to 20 JSON/CSV rows serially; batches stop on a host error and do not promise rollback.
• preview_mogrt_library_publish / publish_mogrt_to_library write a new, immutable v001, v002, … copy under an already-existing local library.
• inspect_after_effects_template_source returns source-comp dimensions, duration, fonts, layer kinds, and controller names on AE 16.1+, while inspect_after_effects_render_templates lists host template names from an existing queue item.
• preview_after_effects_render / enqueue_after_effects_render queue one exact render but never start it. preview_mogrt_premiere_handoff / apply_mogrt_premiere_handoff import into an explicitly named empty MOGRT Verify - … Premiere sequence and read back insertion/control descriptors.

The authoring connector uses AFTER_EFFECTS_MCP_TEMP_DIR (default: the OS temp directory plus after-effects-mcp-bridge), completely separate from PREMIERE_TEMP_DIR. The tool will not create, replace, or switch projects; it requires the already-open AE project and output directory to be contained by the same approved workspace. An import/control readback still needs rendered-frame review before delivery; use capture_frame or a separate approved export path.

inspect_sequence_review_report creates one read-only handoff report from Premiere timeline readback: sequence structure, primary-track gaps, disabled clips, muted tracks, marker timing, and offline-source evidence. It never returns media paths; marker comments are omitted unless explicitly requested. The report is not proof of rendered pixels, audio quality, caption accuracy, rights, or editorial approval.

The MCP handshake reads serverInfo.version from the installed package.json, so clients receive the package version that is actually running rather than a separately maintained literal.

Tools with mixed execution boundaries can provide explicit operational metadata at registration. This is used for local file verification, static feature-support reports, and hybrid local-plus-Premiere validation so the capability catalog does not infer a host dependency from naming alone.

Collaboration and AI feature boundaries

get_advanced_feature_support returns a machine-readable matrix for Productions, Team Projects, Frame.io, Media Intelligence, Generative Extend, Object Mask, caption translation, Speech-to-Text, Enhance Speech, Remix, editorial plans, Premiere AI Assistant, Generative Media, and a future local semantic index. Each entry includes an explicit access mode: direct, observable-only, artifact-import, external-provider, user-assisted, unavailable, or planned. Pass an optional Premiere version, intended backend, confirmed entitlements, and network state to evaluate prerequisites without conflating them with API availability. It distinguishes documented APIs from entitlements, network prerequisites, separate service APIs, and user-assisted operations without using menu automation or private APIs.

The report tool itself is local: it does not contact Premiere and is callable through the current MCP server. Each feature entry separately reports whether its operations are callable through the production CEP transport. Productions reports only static backend/version eligibility until a UXP host performs live capability negotiation.

• Productions exposes documented read-only state through UXP, but the production MCP transport is still CEP.
• Frame.io needs a separately authenticated Frame.io API integration; an account entitlement alone does not make it callable through Premiere's DOM.
• Transcript JSON import/export is documented in UXP. Starting Speech-to-Text is not.
• preview_transcript_edit_uxp and plan_transcript_rough_cut_uxp provide a revision-locked transcript-edit workflow. The planner maps selected transcript ranges to verified 1x placements in a duplicate sequence and emits descending split/remove instructions; it does not claim that Adobe exposes native transcript text deletion or perform an unverified destructive edit.
• The remaining AI operations are user-assisted or unsupported by documented public APIs. The tool explains what can be inspected after a user completes the operation and where artifact provenance cannot be established safely.
• The server never uses menu automation, private APIs, clip-name heuristics, or duration changes as proof that an AI operation occurred.

Reusable project context

For projects where repeatedly inspecting clips, transcripts, audio, and timeline placements is expensive, call manage_project_context with action: "capture". The local context engine indexes a bounded active-sequence snapshot, hashes native project/media paths before persistence, and returns independent source, timeline, and combined context revisions. Add transcript passages, shot descriptions, audio observations, or editor notes once with action: "enrich"; ordinary trims and moves update the timeline revision without discarding unchanged source analysis.

Use search_project_context to retrieve only evidence relevant to the current editing intent. create_context_edit_plan returns a non-mutating candidate scaffold and stale-state guards; create_editorial_plan adds reviewed organization, stringout, rough-cut, and caption-artifact routes without calling an LLM or changing Premiere. Exact identities must still be resolved and passed through the routed operation's preview/confirmation flow before any application. See the project context engine guide for storage controls, privacy boundaries, and invalidation behavior; see local-first AI editorial workflows for the complete review-and-route workflow.

Authenticated UXP connection

The MCP server can accept a local UXP panel connection and invoke the UXP commands that are currently implemented:

PREMIERE_UXP_TOKEN="replace-with-a-long-random-secret" premiere-pro-mcp

Generate that secret yourself and enter the same value in the UXP panel (Window → UXP Plugins, not the CEP Connector). CEP-only setups do not use a token. The listener binds only to 127.0.0.1:7777, authenticates the WebSocket upgrade, requires a versioned capability handshake, correlates concurrent requests, and fails pending work on timeout or disconnect. Set PREMIERE_UXP_PORT to use another loopback port.

When enabled, MCP discovery includes 91 capability-gated UXP additions. The first expansion covers effects, deterministic timeline selection, selection batches, scene detection, proxy/ingest, relink, metadata, color conformance, read-only Premiere/After Effects environment inspection, Source Monitor audition, storage, and least-privilege workspace access. The second adds Project-panel selection, marker CRUD, single-transaction undoable beat-grid marker application, bin organization, sequence settings, imports, typed effect parameters/keyframes, track-item transforms, atomic J/L split edits, SequenceEditor timeline edits, sequence lifecycle, and AME encoding. The third wave begins with a redacted event journal, conservative AME terminal receipts, explicit host-readiness gates, safe multi-project sessions, lease-based growing-media control, namespaced workflow checkpoints, bounded media-health maintenance, caption-aware track mute state, transactional source trim/framing, guarded sequence-range updates, guarded source-media start timing, and guarded source-media frame-rate/pixel-aspect overrides documented in the third-wave workflow matrix. The bounded migration surface also includes a non-ripple selected-item lift, native video-transition listing and guarded transactions, native active and explicit-GUID sequence timing inspection, opt-in installed-MOGRT directory inspection without filesystem enumeration, bounded native timeline structure inspection with opt-in source IDs, classification, and broad content category, guarded sequence display-format updates, a capped native Project-panel tree, a double-read Project-panel insertion-bin snapshot, guarded empty/default sequence creation, bounded native Project-panel schema/item-column metadata inspection, guarded direct Project-panel metadata replacements with exact state/readback guards, guarded typed Project metadata schema-field creation with non-field-level readback, guarded direct updates to three named application preferences, guarded transcript JSON replacement for one exact source clip, direct active-track-item identity readback, guarded source-only slips, guarded contiguous three-item slides, guarded append-only timeline duplicates, guarded contiguous same-track ripple deletes, guarded source-project-item color labels resolved from one timeline coordinate, guarded explicit-sequence preview-frame rectangle updates, explicit opt-in source-media provenance paths with bounded double resolution, bounded source-proxy readiness with an explicit attached-path disclosure, guarded static effect-parameter PointF x/y updates, bounded effect-parameter descriptor catalogs, bounded project/sequence Object Mask audits, native FrameRate/TickTime frame-alignment inspection with caller-owned inputs and tick readback, and native TickTime arithmetic over caller-owned canonical tick strings; it does not claim direct empty-track create/delete, global redo support, playback proof, Object Mask counts or visual validity, an atomic Project-panel metadata compare-and-set, app-preference Undo, installed-template availability or compatibility, imported transcript persistence, override-presence/clear semantics, path existence, source lineage or rights, proxy compatibility, general or keyframed effect-parameter value readback, rendered-frame correctness, linked-item sync, or licensed-host validation. A separate hybrid benchmark gate keeps native acceleration disabled until reproducible cross-platform evidence exists. See also the first stable workflow matrix, the next-ten workflow matrix, the guarded-slip workflow notes, the guarded-slide workflow notes, the guarded-duplicate workflow notes, and the guarded-ripple-delete workflow notes. Commands are advertised only while the authenticated local UXP bridge is connected; the host capability handshake remains the authority for support in the running Premiere build. A failed UXP command is never silently retried through CEP because the first operation may have partially succeeded.

The bounded UXP migration surface also exposes a read-only animated-PointF endpoint-displacement inspection through automate_effect_parameters_uxp. It requires one exact component and parameter target plus a strictly increasing time interval, double-reads native endpoint values and their distance, and refuses static, malformed, or changed parameter state. It does not modify Premiere or prove rendered appearance, playback, persistence, Undo, or licensed-host behavior.

The bounded UXP migration surface also exposes guarded static Color effect-parameter RGBA updates through automate_effect_parameters_uxp. They require a complete double-read snapshot, explicit confirmation, a valid operation ID, per-parameter serialization, one native transaction, and exact raw-component readback. They reject time-varying parameters and do not claim keyframed Color edits, color management, rendered appearance, playback, persistence, Undo, or licensed-host validation.

The bounded UXP migration surface also exposes guarded typed Project metadata schema-field creation. It requires the exact inspected active-project identity and 12 KiB-bounded panel XML, an explicit confirmation and operation ID, and serializes this bridge's schema and panel-replacement calls per project. Adobe provides no atomic compare-and-set or field-level schema getter: native acceptance and a changed panel XML are evidence only, so the command always returns committed_unverified and does not claim field presence, persistence, UI results, Undo, cancellation, or licensed-host validation.

The panel now requests access to one operator-selected workspace instead of declaring full filesystem access. Choose the folder in the panel before invoking a path-based UXP workflow. Media, relink, preset, export, and Source Monitor file paths must remain inside it; the persistent capability token and native root path are never returned over MCP. After a folder is granted, path-based commands resolve each native path through that folder's Entry tree and compare the host nativePath against the approved root, so symlink or junction targets outside the workspace still fail closed. If the host Entry cannot be walked and no resolver is supplied, those commands stay unsupported and the CEP fallback remains the path-based compatibility route.

Native transcript editing starts with a read-only, revision-locked planning flow. Use get_clip_transcript_uxp to export the transcript Premiere generated for a source clip, select source-time ranges from that JSON, and pass its SHA-256 revision to preview_transcript_edit_uxp. The preview sorts and merges ranges and returns a confirmation token without changing the timeline. Premiere does not expose a documented operation that directly turns deleted transcript text into timeline cuts, so automatic application remains withheld until the source-to-sequence mapping and documented reconstruction path pass live-host validation. search_clip_transcript_uxp provides read-only discovery without substituting an external transcription engine.

Premiere 26.2-26.3 hosts also expose documented UXP workflows for revisioned project inspection, verified project saves, preset-based sequence creation, OTIO/FCP XML interchange, transcript-language discovery, Object Mask detection, Adobe Media Encoder control, track renaming, subclip creation, stable marker inspection, Source Monitor positioning, and clip transcript detection. Mutations accept optional idempotency keys and return explicit verification outcomes. See the Adobe UXP 26.3 coverage matrix and the UXP capability foundation for the command matrix and live-host validation boundary.

The Premiere surface registry separately tracks the Premiere DOM, general UXP JavaScript, HTML/CSS, Spectrum, plugin guides, UXP Hybrid C++, the standalone Premiere C++ PrSDK, CEP/ExtendScript, the CEP platform, QE, and pinned competitor sources. The exhaustive general UXP JavaScript inventory is generated from Adobe's pinned declarations, while the exhaustive Premiere documentation inventory tracks every page in Adobe's live sitemap. This keeps declaration and documentation inventory distinct from implementation and licensed-host coverage.

The stable workflow expansion adds native component-chain effects, deterministic timeline selection, compound selection batches, scene-edit detection, proxy/ingest control, guarded offline relink, transactional project/XMP metadata, color and footage-conformance preflight, full Source Monitor audition, and project/Production storage checks. See the stable UXP workflow matrix for exact argument, undo, confirmation, and live-host boundaries.

Architecture

Local (stdio):

┌───────────────┐   stdio (MCP)    ┌──────────────┐   File-based IPC   ┌───────────────┐
│  AI Client    │ ◄──────────────► │  MCP Server  │ ◄────────────────► │  CEP Plugin   │
│  (Claude,     │                  │  (Node.js /  │   .jsx commands    │  (runs inside │
│   Windsurf,   │                  │  TypeScript) │   .json responses  │  Premiere)    │
│   Cursor,     │                  └──────────────┘                    └──────┬────────┘
│   Copilot)    │                                                             │
└───────────────┘                                                             │ evalScript()
                                                                              ▼
                                                                       ┌───────────────┐
                                                                       │  Premiere Pro │
                                                                       │  ExtendScript │
                                                                       │  + QE DOM     │
                                                                       └───────────────┘

Remote (HTTP/SSE — Fly.io):

┌───────────────┐  HTTP+SSE (MCP)  ┌─────────────────────┐   File-based IPC   ┌──────────────┐
│  AI Client    │ ◄──────────────► │  MCP Server         │ ◄────────────────► │  CEP Plugin  │
│  (any MCP     │                  │  premiere-pro-mcp   │   .jsx / .json     │  (Premiere)  │
│   client)     │                  │  .fly.dev           │   shared volume    └──────────────┘
└───────────────┘                  └─────────────────────┘

1. AI client invokes an MCP tool (e.g., add_to_timeline)
2. MCP server generates ES3-compatible ExtendScript with helper functions prepended
3. Script is written to a .jsx command file in a shared temp directory
4. CEP plugin polls for command files, executes via CSInterface.evalScript()
5. Result JSON is written to a response file and returned to the AI

The file-based IPC bridge is simple, reliable, and works across macOS and Windows without network sockets.

inspect_unique_object_identity_uxp is a separate read-only native identity inspection route. It resolves exactly one project item or sequence, reads the opaque UniqueSerializeable identity twice, and rejects drift without retaining the value or treating it as edit authority. See the unique-identity workflow notes for its bounds and proof boundary.

Film editorial review

inspect_film_editorial_workflow validates captured identities and explicit source/scene coverage, builds marker/stringout/line/beat review artifacts, and reports revision-bound notes, VFX state, change impact and turnover exceptions. It performs local inspection; host edits and exports use separate guarded tools. See usage, example and remaining execution adapters.

Reviewed assistant-editing tools now provide dialogue candidate decisions, quote paper-edit CSV, batch copy previews, sync-evidence triage, and owned-footage B-roll review. These are local evidence reviews, not new host engines. See review workflow contracts.

Tools (393 core total; 391 under the default profile; 487 with a connected UXP bridge)

The complete supported-actions catalog lists every registered core tool, the two tools restricted behind explicit unsafe-script authority, and all 96 authenticated UXP additions with their current action or mode values. It is generated from the same MCP registration surface returned to clients; the tables below are a shorter workflow-oriented overview.

Discovery & Inspection (10 + 10)

Project Management (26)

Timeline & Editing (11 + 27 advanced)

Premiere Pro 26.3 compatibility: some installations silently ignore QE structural edits (ripple_delete, razor/split) and existing effect-parameter writes. These tools now verify the resulting sequence state and return an error instead of a false success. For structural edits, rebuild the wanted source ranges into a new sequence with create_sequence and add_to_timeline. The legacy CEP transition path targets qeClip.addTransition (not the QE track) and reports success only after DOM transition-count readback. Prefer the connected UXP transition tools where available; overlay clips remain a workaround when a legacy QE write is rejected or not verified. See issue #21.

Speed, caption, and visual-keyframe boundaries: Premiere Pro 26.3 may reflect legacy QE speed/direction methods, but exposes no supported scripting setter or Time Remapping component for timeline-clip speed or direction (documented UXP through 26.3 has only getSpeed / isSpeedReversed). reverse_clip, speed_change, set_clip_speed_qe, and set_clip_properties with speed now stop before host mutation; use set_clip_duration to change timeline length, or the Speed/Duration UI or pre-rendered media to retime. add_text_overlay likewise stops before mutation because a raw-text-to-caption API is not exposed; import an .srt/.vtt and use create_caption_track, or use a MOGRT/PNG overlay. Keyframe and caption-track responses can prove parameter/structure readback only—not rendered pixels—so verify playback or exported frames before delivery. On macOS, AME preset discovery scans each installed app bundle's Contents/MediaIO/systempresets; prefer a Match Source preset for vertical projects.

Host limits: The capability ledger distinguishes unavailable APIs, unverified host requests, and property/structure readback. A verified parameter or component does not establish rendered appearance.

Verified track edits: add_track and add_tracks validate requested counts and return success only when the active sequence's track counts exactly match the request. overwrite_clip validates both selected track indices and confirms the requested source item appears at the requested frame on a target track. On a Premiere 26.x build that ignores any of these calls, the MCP response is an error with the observed state rather than a false success. These are automated CEP contracts, not proof of a particular licensed host configuration.

trim_clip accepts exactly one source-relative new_in_seconds or new_out_seconds per call. It refuses retimed clips because CEP cannot prove their source-to-timeline mapping, then reads both source points and visible timeline start/end/duration before reporting success. The default keyframe_policy: "reject" stops before a trim that would leave effect keyframes beyond the visible clip; keyframe_policy: "preserve" is an explicit opt-in and reports the remaining count. split_clip verifies a spanning clip, the expected count increase, and the left/right cut boundaries. Its QE path cannot prove effect-keyframe redistribution, so a successful result labels those semantics unverified. These are CEP contract checks, not validation in a licensed Premiere Pro 26.x host.

set_clip_duration changes a placed clip's timeline length without touching source in/out directly: pass exactly one of duration_seconds (from the current start) or end_seconds (absolute timeline end). It writes the documented TrackItem.end as a tick-based Time, refuses an end that would overlap the next clip on the same track, applies the same keyframe_policy guard as trim_clip when shortening, and reads start/end back. If Premiere clamps the end (for example, video media with no remaining handle, or a still whose project item has in/out points), the original end is restored and the tool returns an error. Linked audio/video partners are not adjusted. This is a CEP contract check, not validation in a licensed Premiere Pro host.

Effects & Color (8)

Premiere 26.x component removal: remove_effect and remove_effect_by_name require the CEP Component.remove() method. Some 26.x components, including Essential Sound's Amplify, do not expose that method. The tools return an actionable capability error and leave the component unchanged; use Effect Controls to remove it manually. The QE DOM has no safe targeted-removal fallback.

Essential Sound audio automation: Essential Sound can write ducking or level automation to an Amplify component rather than the clip's Volume > Level. adjust_audio_levels, set_clip_volume, and get_clip_volume operate only on Volume > Level, so they do not read, change, or verify Amplify automation. Inspect the clip's components (or Effect Controls) before treating a Volume readback as the clip's final gain.

Keyframes (8)

Export & Encoding (17)

Premiere's documented automation surfaces do not currently expose OTIO or native EDL interchange, Render and Replace, cloud publishing, or Content Credentials export configuration. export_sequence_edl therefore builds the CMX 3600 list locally from the clips Premiere reads back and validates it with the same parser used by inspect_cmx3600_edl; it is not a Premiere-native export. get_capabilities reports the remaining delivery gaps explicitly rather than presenting UI-only operations as available tools.

Source Monitor & Playback (7 + 4)

Selection & Clipboard (7 + 7)

Media Properties (16)

Sequence Management (11)

Workspace & Captions (2 + 1)

Timeline QA (2)

Editor Requests: markers, selection, checkpoints, navigation (5)

Review and Conversation Planning (2)

Both planners are local and deterministic. Premiere exposes no scripting API for switching angles inside a multicam source sequence, so the multicam plan targets synced clips stacked on separate video tracks and is applied through razor_all_tracks, select_clips_in_range, batch_enable_disable, and add_markers_batch on a checkpointed sequence. See editor requests for the community and competitor evidence behind these tools and their verification boundaries.

Speaker Layout (2)

Mask Fit (1)

Rhythm Plans (2)

Shorts Intelligence (2)

Caption Authoring (2)

Reaction Shorts (3)

Transcript Word Edits (4)

Platform Delivery Planning (2)

Scripting (2)

...and 100+ more

Track targeting, batch operations, markers, audio levels, motion/transform, metadata, sequence settings, navigation, project analysis, and more. Run get_project_info to get started — the AI will discover what it needs.

MCP Resources

The server exposes fourteen LLM context resources and nineteen workflow prompts:

The ten premiere:// snapshots are read-only CEP bridge requests. They include a revision token for stale-state detection and omit native media, project-tree, preset, and output paths. A successful snapshot proves bridge readback only—not licensed-host feature coverage, playback, rendering, or editorial correctness.

Remote Deployment (Fly.io)

The server includes an HTTP/SSE transport (src/http-server.ts) for remote access via mcp-remote or any MCP client that supports Streamable HTTP.

An operator-managed MCP endpoint runs at https://premiere-pro-mcp.fly.dev/mcp (health check: /health). The website is https://premiere-pro-mcp.com, served from a separate repository; other paths on the Fly host redirect there. The endpoint is not a public desktop relay: it cannot connect an authenticated user to Premiere on that user's computer. Public users should use the local stdio setup until the separate device-pairing relay is available.

Connect to an operator-managed instance

{
  "mcpServers": {
    "premiere-pro": {
      "command": "npx",
      "args": ["mcp-remote", "https://your-authorized-instance.example/mcp"]
    }
  }
}

The instance must either provision an operator bearer token or use the OAuth resource-server configuration below. The production endpoint intentionally returns 401 to callers who have not been authorized.

Self-host on Fly.io

# Clone and deploy your own instance
git clone https://github.com/leancoderkavy/premiere-pro-mcp.git
cd premiere-pro-mcp
fly apps create your-app-name
# Required: add bearer token auth. Use a unique, high-entropy secret per deployment.
fly secrets set MCP_AUTH_TOKEN=your-secret-token
fly deploy --remote-only

Then connect with:

{
  "mcpServers": {
    "premiere-pro": {
      "command": "npx",
      "args": ["mcp-remote", "https://your-app-name.fly.dev/mcp",
               "--header", "Authorization: Bearer your-secret-token"]
    }
  }
}

Trusted-operator OAuth resource-server mode

For an identity-aware operator deployment, configure a real OAuth/OIDC authorization server rather than distributing MCP_AUTH_TOKEN. The authorization server must support the MCP client's registration model and issue signed access tokens with an exact audience for this MCP resource.

fly secrets set \
  MCP_OAUTH_ISSUER=https://identity.example.com \
  MCP_OAUTH_JWKS_URI=https://identity.example.com/.well-known/jwks.json \
  MCP_OAUTH_AUDIENCE=https://your-app-name.fly.dev/mcp \
  MCP_PUBLIC_URL=https://your-app-name.fly.dev \
  MCP_OAUTH_REQUIRED_SCOPES=premiere:mcp \
  MCP_OAUTH_ALLOWED_SUBJECTS=your-provider-user-subject

OAuth mode validates the token signature, algorithm, issuer, exact audience, expiry, issued-at time, subject, and required scopes. It publishes protected resource metadata at /.well-known/oauth-protected-resource/mcp and includes that URL in the WWW-Authenticate challenge. Configuration is fail-closed: partial OAuth settings, non-HTTPS production URLs, ambiguous OAuth/shared-token settings, and missing credentials all prevent startup.

MCP_OAUTH_ALLOWED_SUBJECTS is mandatory and restricts this single-bridge deployment to explicitly trusted operator identities. This is an enforcement boundary, not a public-user device model.

This mode authenticates trusted operators but does not yet implement device ownership, desktop pairing, or per-user Premiere routing. Do not expose editor mutations as a public multi-user service until an outbound desktop relay and durable user/device authorization are implemented.

Note: The file bridge still requires the CEP plugin to share the same PREMIERE_TEMP_DIR. For cloud deployments this means running a sync agent or using fly proxy / WireGuard to reach your local machine. detect_silence can analyze only media paths available inside the server filesystem; a desktop-only path is not automatically available to a remote Fly machine. For a shared or multi-user remote deployment, put a managed identity-aware edge in front of the server and replace the shared bearer secret with per-user authorization. The built-in limiter is intentionally process-local defense in depth, not a substitute for an edge/WAF or account system.

Environment Variables

When PostHog is enabled, the server records mcp_connection_attempt, mcp_request, mcp_request_rejected, and mcp_tool_call. It also records premiere_mcp_activation_completed only after the read-only verify_premiere_connection check confirms the selected bridge, an open project, and an active sequence. Events contain bounded operational fields such as method, tool name, outcome, status code, duration, and selected bridge. Authentication tokens, IP addresses, MCP arguments, project paths, media names, and tool results are never sent. Person profiles are disabled for these events.

This signal is an aggregate emission, not proof that an analytics provider received it, a count of unique people or editors, or evidence that an editing workflow succeeded. It does not carry a person or editor identifier, so it cannot safely infer "first value."

Project Structure

premiere-pro-mcp/
├── src/
│   ├── index.ts                 # Entry point — stdio transport setup
│   ├── http-server.ts           # Entry point — HTTP/SSE transport (Fly.io / remote)
│   ├── server.ts                # MCP server — registers 393 tools, filtered by authority profile
│   ├── bridge/
│   │   ├── file-bridge.ts       # File-based IPC (write .jsx, poll .json)
│   │   └── script-builder.ts    # ExtendScript generator with ES3 helpers
│   ├── tools/                   # 51 tool modules
│   │   ├── discovery.ts         # Project discovery and queries
│   │   ├── recovery.ts          # Read-only autosave discovery and private bridge telemetry
│   │   ├── project.ts           # Project management and import
│   │   ├── media.ts             # Media and proxy management
│   │   ├── sequence.ts          # Sequence creation and settings
│   │   ├── timeline.ts          # Timeline clip operations
│   │   ├── effects.ts           # Effect application and color correction
│   │   ├── transitions.ts       # Transition management (QE DOM)
│   │   ├── audio.ts             # Audio levels, keyframes, and ffmpeg silence analysis
│   │   ├── av-settings.ts       # Documented AV inspection, mapping, and capability boundaries
│   │   ├── text.ts              # Text overlays and MOGRTs
│   │   ├── markers.ts           # Sequence and clip markers
│   │   ├── tracks.ts            # Track add/delete/lock/visibility
│   │   ├── playhead.ts          # Playhead, work area, in/out points
│   │   ├── metadata.ts          # Metadata, XMP, color labels
│   │   ├── export.ts            # Export, frame capture, encoding
│   │   ├── advanced.ts          # QE DOM: ripple, roll, slide, slip, speed
│   │   ├── keyframes.ts         # Keyframe CRUD and interpolation
│   │   ├── scripting.ts         # Execute arbitrary ExtendScript
│   │   ├── inspection.ts        # Deep project/sequence/clip inspection
│   │   ├── selection.ts         # Clip selection utilities
│   │   ├── clipboard.ts         # Copy effects, batch operations
│   │   ├── source-monitor.ts    # Source monitor control
│   │   ├── track-targeting.ts   # Track targeting, motion, audio props
│   │   ├── utility.ts           # Batch ops, analysis, navigation
│   │   ├── health.ts            # Connectivity ping
│   │   ├── workspace.ts         # Workspace layout switching
│   │   ├── captions.ts          # Caption track creation
│   │   ├── playback.ts          # Timeline/source playback control
│   │   └── project-manager.ts   # Project consolidation/transfer
│   └── resources/
│       └── extendscript-reference.ts  # API reference for LLM context
├── cep-plugin/                  # CEP panel that runs inside Premiere Pro
│   ├── CSXS/manifest.xml        # Extension manifest (PPRO 14.0+)
│   ├── index.html               # Panel UI
│   ├── main.js                  # Bridge polling and script execution
│   ├── host.jsx                 # ExtendScript entry point
│   └── CSInterface.js           # Adobe CEP interface library
├── after-effects-cep-plugin/    # Separate AE CEP bridge for guarded MOGRT recipes
├── scripts/
│   ├── install-cep.sh           # macOS CEP installer (symlink + debug mode)
│   └── install-cep.ps1          # Windows CEP installer (copy + REG_SZ debug mode)
├── Dockerfile                   # Multi-stage Docker build for Fly.io
├── fly.toml                     # Fly.io deployment config
├── RESEARCH.md                  # API research and implementation status
├── AGENTS.md                    # IDE / coding-agent map
├── CONTRIBUTING.md              # Contribution guidelines
├── CHANGELOG.md                 # Version history
└── LICENSE                      # MIT License

Technical Details

CEP and UXP backends

CEP remains the production backend because it provides broad ExtendScript access and the undocumented QE DOM used for effects, ripple deletes, and advanced trims across Premiere Pro 2020–2026. The packaged uxp-plugin is a Premiere 25.6+ preview backend for supported frame export, capability discovery, and state events. It does not silently retry failed UXP mutations through CEP.

ExtendScript Compatibility

All generated scripts use ES3 syntax (var, manual for loops, no arrow functions, no let/const) since ExtendScript is based on ECMAScript 3. The bridge writes a versioned helper library to the shared temp directory and loads it once per ExtendScript engine via $.evalFile; each command then sends only its tool-specific script.

Security

Metadata privacy: CEP get_metadata and get_xmp_metadata now default to bounded parsed fields with GPS, serial, author and contact values omitted. Raw packets require explicit packet flags and include_sensitive: true; source media paths require include_media_path: true. Existing callers expecting raw XML must opt in. Disk context can retain names, transcripts and notes; select PREMIERE_CONTEXT_BACKEND=memory to avoid persistence.

Understand the trust model before deploying this: any client that can reach the MCP server can control Premiere Pro. execute_extendscript and evaluate_expression are arbitrary-code-execution tools by design and are omitted from discovery and denied at call time by default. Enable them only by setting PREMIERE_MCP_CAPABILITIES=inspect,edit,export,filesystem,unsafe-script.

• Run it locally over stdio unless you have a specific reason not to. That's the safe default.
• The HTTP transport (http-server) requires MCP_AUTH_TOKEN and refuses to start without it in production. It binds 0.0.0.0 and is remotely reachable, so never expose it publicly without a strong token and edge controls. ALLOW_UNAUTHENTICATED=1 is limited to non-production local/test use.
• The HTTP transport admits only exact /mcp Streamable HTTP requests, enforces body/socket/request limits, and applies a bounded in-process per-credential rate and concurrency limit before MCP request parsing or Premiere bridge work begins. It returns 413, 429, or 503 on containment failures. Configure an upstream rate limit and request-size limit too; process-local counters do not protect a multi-machine deployment.
• The HTTP transport serves no web pages. It answers only /mcp, /health, and OAuth protected-resource metadata under a deny-all Content Security Policy. Other GET/HEAD paths on premiere-pro-mcp.fly.dev or a *.premiere-pro-mcp.com host redirect (308) to https://premiere-pro-mcp.com; every other host receives a JSON 404.
• Media scans yield between asynchronous filesystem operations and bound total traversal: 25,000 entries, 5,000 matching files, 2,000 directories, depth 32, a 1,000-directory pending queue, and a cooperative five-second budget. A stalled OS call can exceed that elapsed budget. Inspect scan_incomplete and scan_limit_reasons before treating a watch baseline as complete; import previews also report incomplete.
• Project backups stream their copy and checksum work with a configurable byte budget. Only one backup runs per process; concurrent requests fail promptly. Failed or cancelled copies are removed without replacing an existing backup.
• The Docker runner uses the unprivileged node user (UID 1000). Custom bridge or context volume mounts must be writable by that user and private to the operator. The default context directory is under /home/node/.local/state/premiere-pro-mcp.
• Dependency audits run in CI for both lockfiles. Keep local .env*, .npmrc, and private key/certificate files out of commits and Docker build contexts; only placeholder .env.example and .env.template files are eligible for Git tracking.
• Both CEP panels validate the bridge directory before writing a heartbeat or polling commands. Symlinks and directories owned by another user are refused. On POSIX, directories writable by group/other users are refused by both the server and panels; tightening permissions alone cannot make previously staged commands trustworthy. On Windows, only the current user, SYSTEM and Administrators may have write access, including inherited grants. Windows capability and app-container SIDs (S-1-15-*) are ignored: they are not logon principals and appear as inherited FullControl on stock %LOCALAPPDATA% paths. ACL inspection failures also stop startup. Ancestors must also prevent other users from replacing the bridge path; a private child inside a broadly writable parent is insufficient (POSIX sticky temp directories are supported). Choose a new directory under a private user-owned location if the connector rejects a shared temp folder, and configure the same path on the server and panel. Do not copy pending commands from the rejected folder into the new one. Windows checks invoke PowerShell synchronously before publishing commands, with a five-second timeout and 64 KiB output cap; this adds command latency rather than caching a permission decision that could become stale.
• There is a 500 KB script size limit, and a small regex check that rejects eval(), new Function(), and System.callSystem() in tool-generated scripts. This is a guard rail, not a sandbox — it is trivially bypassable and is not a security boundary. Do not rely on it to contain untrusted input; the real boundary is who can reach the server.

QE DOM

Many tools use the undocumented QE DOM (enabled via app.enableQE()). These tools are marked with "Uses QE DOM" in their descriptions. The QE DOM provides capabilities unavailable through the standard ExtendScript API:

• Apply effects and transitions by name
• Ripple delete, roll/slide/slip edits
• Set clip speed and reverse
• Frame blending and time interpolation
• Remove all effects from a clip

Frequently asked questions

What is an MCP server for Adobe Premiere Pro? It is a Model Context Protocol server that gives a compatible AI assistant structured, reviewable control over supported Adobe Premiere Pro workflows on your own computer. This project's display name is MCP for Adobe Premiere Pro; its npm package is premiere-pro-mcp.

How do I use it with Claude? Install the Claude Desktop bundle, install the separate signed Premiere connector, restart Premiere, then ask Claude to run verify_premiere_connection with no changes. That first prompt is read-only. Claude Fable 5.1 is an optional client model for Cursor, Claude Desktop, or Claude Code; see Claude Fable 5.1 workflows.

Where is the GitHub repository? https://github.com/leancoderkavy/premiere-pro-mcp. Releases, the signed .zxp connector, and the Claude .mcpb bundle are on that repository's Releases page.

What is the bridge panel in Premiere Pro? A local CEP panel (Window > Extensions > MCP for Adobe Premiere Pro) that connects the server to a running Premiere Pro instance. "Running" in the panel means the local bridge is available; it does not by itself show that an edit completed. An authenticated UXP panel is an additional, capability-gated route.

Is there a setup guide? Yes - premiere-mcp-setup-guide.md is a portable guide you can attach to any AI assistant, and https://premiere-pro-mcp.com/docs/ hosts the published documentation.

Does it run on Windows and macOS? Both. Node.js 20.19 or newer and Premiere Pro 2020-2026 are required, and the assistant, server, connector, and Premiere should stay on the same computer. Individual tool support remains capability- and host-dependent.

Is this Adobe's AI Assistant? No. This is an independent MIT-licensed project, not an Adobe product and not Adobe's native AI Assistant. It is also distinct from other MCP servers for Premiere Pro; premiere-pro-mcp is the only npm package published from this repository.

Does local-first mean my AI assistant sends no data to a cloud service? The recommended Premiere bridge runs locally. Your AI client's handling of prompts and tool results still applies. Review its settings before sharing sensitive project information; local-first is not a universal no-upload guarantee.

Is every action previewed, undoable, and host-verified? Tool contracts differ. Review action modes and approval requirements before calling an edit. Readback-confirmed results are verified; committed_unverified means the host accepted a change but verification is incomplete. Inspect Premiere before retrying an uncertain mutation. Save a duplicate project for evaluation and confirm rendered output separately.

Which Adobe Premiere Pro tasks are good first automation candidates? Start with a read-only project and timeline inventory. Then evaluate one bounded workflow on a duplicate project: approved media imports, bin organization, marker preparation, or review-frame exports. The automation guide links exact tool contracts and explains what evidence to check. A still-frame export does not prove playback or audio; a local delivery QC scan does not prove every platform requirement.

Where can search engines and AI assistants find current product facts? Use the website's facts page and generated AI reference, which describe the integrity-verified published npm artifact. The source counts in this README can include changes not yet published. Neither catalog counts nor unit tests establish that an operation succeeded on a particular Premiere host.

Is it free? Yes. The server, the CEP connector, and the documentation are MIT licensed and free to use.

Troubleshooting