Use a Gateway - Zed

Use a gateway when you route model requests through a platform such as OpenRouter, Vercel AI Gateway, Amazon Bedrock, or another OpenAI-compatible service.

OpenRouter

Use OpenRouter when you want to route Zed AI features through OpenRouter.

1. Visit OpenRouter and create an account.
2. Generate an API key from your OpenRouter keys page.
3. Open Settings → AI → LLM Providers with agent::OpenSettings and find the OpenRouter row.
4. Enter your OpenRouter API key.

Zed also reads OPENROUTER_API_KEY from the local Zed process environment.

When using OpenRouter as your assistant provider, explicitly select a model in your settings:

{
  "agent": {
    "default_model": {
      "provider": "openrouter",
      "model": "openrouter/auto"
    }
  }
}

The openrouter/auto model routes requests to an available model selected by OpenRouter. You can also specify any model available through OpenRouter's API.

OpenRouter Custom Models

You can add custom models to the OpenRouter provider in settings:

{
  "language_models": {
    "open_router": {
      "api_url": "https://openrouter.ai/api/v1",
      "available_models": [
        {
          "name": "google/gemini-2.0-flash-thinking-exp",
          "display_name": "Gemini 2.0 Flash (Thinking)",
          "max_tokens": 200000,
          "max_output_tokens": 8192,
          "supports_tools": true,
          "supports_images": true,
          "mode": {
            "type": "thinking",
            "budget_tokens": 8000
          }
        }
      ]
    }
  }
}

Custom model entries support fields such as name, display_name, max_tokens, max_output_tokens, max_completion_tokens, supports_tools, supports_images, and mode.

OpenRouter Provider Routing

You can control how OpenRouter routes a custom model request among upstream providers with the provider object on each model entry.

Supported fields include order, allow_fallbacks, require_parameters, data_collection, only, ignore, quantizations, and sort.

{
  "language_models": {
    "open_router": {
      "available_models": [
        {
          "name": "openrouter/auto",
          "display_name": "Auto Router",
          "max_tokens": 2000000,
          "supports_tools": true,
          "provider": {
            "order": ["anthropic", "openai"],
            "allow_fallbacks": true,
            "require_parameters": true,
            "data_collection": "allow"
          }
        }
      ]
    }
  }
}

Vercel AI Gateway

Use Vercel AI Gateway when you want to route Zed AI features through Vercel.

1. Create an API key from your Vercel AI Gateway keys page.
2. Open Settings → AI → LLM Providers with agent::OpenSettings and find the Vercel AI Gateway row.
3. Enter your Vercel AI Gateway API key.

Zed also reads VERCEL_AI_GATEWAY_API_KEY from the local Zed process environment.

You can set a custom endpoint for Vercel AI Gateway in settings:

{
  "language_models": {
    "vercel_ai_gateway": {
      "api_url": "https://ai-gateway.vercel.sh/v1"
    }
  }
}

Amazon Bedrock

Use Amazon Bedrock when you want model access through AWS.

Bedrock supports models that support streaming tool use. See Amazon Bedrock's Tool Use documentation.

Your AWS credentials need these permissions:

• bedrock:InvokeModelWithResponseStream
• bedrock:InvokeModel

Bedrock supports Zed-prefixed AWS environment variables so Zed does not override or consume your normal AWS credentials:

• ZED_ACCESS_KEY_ID
• ZED_SECRET_ACCESS_KEY
• ZED_SESSION_TOKEN
• ZED_AWS_PROFILE
• ZED_AWS_REGION
• ZED_AWS_ENDPOINT
• ZED_BEDROCK_BEARER_TOKEN

Bedrock Authentication

You can authenticate with a named profile, static credentials, or a Bedrock API key.

For a named profile, configure Bedrock in settings:

{
  "language_models": {
    "bedrock": {
      "authentication_method": "named_profile",
      "region": "your-aws-region",
      "profile": "your-profile-name"
    }
  }
}

For static credentials, open Agent Settings with agent::OpenSettings, go to the Amazon Bedrock section, and enter the access key ID, secret access key, and region.

For a Bedrock API key, choose API key authentication:

{
  "language_models": {
    "bedrock": {
      "authentication_method": "api_key",
      "region": "your-aws-region"
    }
  }
}

The API key itself is stored in the system keychain, not in settings.json.

Bedrock Cross-Region Inference

Zed uses Cross-Region inference for Bedrock on a best-effort basis.

By default, Zed uses regional inference profiles. To opt into global profiles, add allow_global:

{
  "language_models": {
    "bedrock": {
      "authentication_method": "named_profile",
      "region": "your-aws-region",
      "profile": "your-profile-name",
      "allow_global": true
    }
  }
}

Only some models support global inference profiles. See the AWS Bedrock supported models documentation for the current list.

Bedrock Guardrails

Some AWS environments require a guardrail on every Bedrock API call. Add guardrail_identifier to apply a guardrail to all Bedrock requests:

{
  "language_models": {
    "bedrock": {
      "guardrail_identifier": "arn:aws:bedrock:us-east-1:123456789012:guardrail/abc123",
      "guardrail_version": "DRAFT"
    }
  }
}

Custom Bedrock Models

Add models that are not yet built into Zed with available_models. Use the Bedrock model ID as name, including a region prefix when you want a specific inference profile:

{
  "language_models": {
    "bedrock": {
      "available_models": [
        {
          "name": "arn:aws:bedrock:us-east-1:123456789012:application-inference-profile/abcdef123456",
          "display_name": "Grok 4.3 (ARN)",
          "max_tokens": 500000,
          "max_output_tokens": 131072,
          "supports_tools": true,
          "supports_images": true,
          "thinking": true
        },
        {
          "name": "us.moonshotai.kimi-k3",
          "display_name": "Kimi K3",
          "max_tokens": 1000000,
          "max_output_tokens": 64000,
          "supports_tools": true,
          "supports_images": true,
          "thinking": {
            "adaptive": true,
            "has_xhigh": true,
            "budget_tokens": 5000
          }
        }
      ]
    }
  }
}

name is sent to Bedrock as the model ID. Use the full ID, including a geo prefix or ARN when the model requires one (for example us.anthropic.claude-sonnet-4-7 or us.xai.grok-4.6). Set supports_tools and supports_images for models that support those features. Set thinking to true to enable thinking, or to an object with "adaptive": true, optional "has_xhigh": true, and optional budget_tokens. Leave default_temperature unset for models that reject the temperature field, such as xAI and MoonshotAI models.

Bedrock Custom Endpoints

To send Bedrock Converse requests to a proxy or gateway instead of AWS, set endpoint_url:

{
  "language_models": {
    "bedrock": {
      "endpoint_url": "https://gateway.example.com/bedrock",
      "region": "us-east-1",
      "authentication_method": "api_key"
    }
  }
}

Mantle models, including the built-in GPT-5.6, GPT-5.5, GPT-5.4, and Grok 4.3 models and any models in mantle_available_models, are called through a different service with a different request shape, so they ignore endpoint_url and go to AWS.

If your gateway serves one of those models over the Converse API, add it as a custom Bedrock model so Zed calls it through endpoint_url. Use the model ID your gateway expects as name:

{
  "language_models": {
    "bedrock": {
      "endpoint_url": "https://gateway.example.com/bedrock",
      "available_models": [
        {
          "name": "us.openai.gpt-5.6-luna",
          "display_name": "GPT-5.6 Luna (Gateway)",
          "max_tokens": 1000000,
          "max_output_tokens": 128000,
          "supports_tools": true,
          "supports_images": true
        }
      ]
    }
  }
}

Don't use an ID that a Mantle model already uses as name, whether it's built in (such as gpt-5.6-luna) or listed in mantle_available_models. A Mantle model with the same ID replaces the custom model, so requests go to AWS instead of your gateway.

Bedrock Mantle Models

Some models, such as the GPT-5.6 family (Sol, Terra, and Luna), GPT-5.5, GPT-5.4, and Grok 4.3, aren't available through Bedrock's Converse API and are only reachable through bedrock-mantle, AWS's OpenAI-compatible inference endpoint. Zed routes these models through bedrock-mantle automatically; they appear alongside the rest of the Bedrock models in the model picker once you're authenticated, with no extra configuration required.

Mantle models require IAM permissions for the bedrock-mantle endpoint (for example via the AmazonBedrockMantleInferenceAccess managed policy) in addition to whatever permissions your existing Bedrock credentials already have, and bedrock-mantle is only available in some AWS Regions. Zed surfaces an error naming the current Region and the supported ones if you try to use a Mantle model outside of them.

Custom Bedrock Mantle Models

You can add custom models served through bedrock-mantle with mantle_available_models:

{
  "language_models": {
    "bedrock": {
      "mantle_available_models": [
        {
          "name": "openai.gpt-oss-120b",
          "display_name": "GPT-OSS 120B",
          "max_tokens": 128000,
          "protocol": "chat_completions",
          "supports_tools": true,
          "supports_images": false,
          "supports_thinking": true
        }
      ]
    }
  }
}

protocol selects which OpenAI-compatible API the model is called through, and must be either chat_completions or responses. Set supports_thinking to true for custom Mantle models that accept OpenAI reasoning effort parameters; Zed will then expose low, medium, high, and xhigh in the thinking effort picker, while disabling thinking sends none.

OpenAI-Compatible Gateways

If your gateway exposes an OpenAI-compatible API, configure it with Use API Access.