mcp-openapi-bridge
Bridges any OpenAPI 3.x REST API to Claude Code by automatically generating one tool per endpoint from your spec, with full argument validation and auth support.
README
mcp-openapi-bridge
A generic MCP server that bridges any OpenAPI 3.x REST API to Claude Code — automatically generating one tool per endpoint from your spec.
Point it at an OpenAPI spec, set your auth token, and Claude can call every endpoint in your API with full argument validation.
Requirements
- Node.js >= 20
How it works
- Loads your OpenAPI 3.x spec from a local file (
OPENAPI_SPEC_PATH) or URL (OPENAPI_SPEC_URL) - Resolves
$refreferences and extracts all operations - Registers one MCP tool per operation (filtered by
OPENAPI_MAX_TOOLS, tags, or path prefix) - Each tool accepts path params, query params, and a
bodyfor request bodies — plus per-callbearer_tokenandcustom_headersoverrides - Executes HTTP requests against
API_BASE_URLand returns the response
Quick start
npm install -g mcp-openapi-bridge
Add to Claude Code:
claude mcp add mcp-openapi-bridge \
-e API_BASE_URL=https://api.example.com \
-e OPENAPI_SPEC_URL=https://api.example.com/openapi.json \
-e OPENAPI_TOKEN=your-token
Or with a local spec file:
claude mcp add mcp-openapi-bridge \
-e API_BASE_URL=https://petstore3.swagger.io/api/v3 \
-e OPENAPI_SPEC_PATH=/path/to/openapi.yaml \
-e OPENAPI_TOKEN=your-token
Environment variables
Required
| Variable | Description |
|---|---|
API_BASE_URL |
Base URL of the REST API, e.g. https://api.stripe.com |
OPENAPI_SPEC_PATH or OPENAPI_SPEC_URL |
Source of the OpenAPI spec (local file or URL) |
Authentication
All auth mechanisms can be combined and are applied simultaneously:
| Variable | Description |
|---|---|
OPENAPI_TOKEN |
Bearer token → Authorization: Bearer <token> |
OPENAPI_API_KEY |
API key value (pair with one of the two below) |
OPENAPI_API_KEY_HEADER |
Header name for API key, e.g. X-API-Key |
OPENAPI_API_KEY_PARAM |
Query param name for API key, e.g. api_key |
OPENAPI_CUSTOM_HEADERS |
Extra headers for every call: X-Foo=bar,X-Baz=qux |
Per-call overrides: Every tool also accepts bearer_token and custom_headers arguments. If provided, they override the global env vars for that single request. This lets Claude switch tokens per call without restarting the server.
Filtering (for large specs)
| Variable | Description | Default |
|---|---|---|
OPENAPI_MAX_TOOLS |
Maximum tools to register | 128 |
OPENAPI_INCLUDE_TAGS |
Comma-separated tags to include | all |
OPENAPI_EXCLUDE_TAGS |
Comma-separated tags to exclude | none |
OPENAPI_PATH_PREFIX |
Only include paths starting with this prefix | all |
For APIs with hundreds of endpoints (Stripe, GitHub, Kubernetes), use these to scope down to the subset you need:
OPENAPI_INCLUDE_TAGS=payments,customers
# or
OPENAPI_PATH_PREFIX=/v1/charges
Transport
| Variable | Description | Default |
|---|---|---|
MCP_TRANSPORT |
stdio or http |
stdio |
PORT |
HTTP port (when MCP_TRANSPORT=http) |
8080 |
MCP_AUTH_TOKEN |
Bearer token required by the /mcp HTTP endpoint. Recommended for any public-routable deployment. |
unauthenticated if unset |
Requests to /mcp are capped at 10MB. See docs/architecture.md
for how stdio and HTTP differ in what's isolated per client, and why API_BASE_URL is
fixed per deployment rather than a per-request parameter.
Tool naming
- If the operation has an
operationId, it's used directly (sanitized to[a-zA-Z0-9_]+) - Otherwise:
{method}_{path}— e.g.GET /orders/{id}/items→get_orders_id_items - Duplicate IDs get a numeric suffix:
getOrders,getOrders_2
Tool arguments
Each tool exposes:
- Path parameters — by name, required
- Query parameters — by name, optional
body— request body for POST/PUT/PATCH (optional or required per spec)- Header parameters — as
header_{lowercase_name} bearer_token— per-call bearer overridecustom_headers— per-call extra headers ({"X-Tenant": "abc"})
Generic fallback tool
execute_rest is always registered for arbitrary requests when no specific tool fits:
execute_rest(method, path, query?, body?, headers?, bearer_token?, custom_headers?)
Docker
Stdio transport:
docker build -t mcp-openapi-bridge .
docker run -i --rm \
-e API_BASE_URL=https://api.example.com \
-e OPENAPI_SPEC_URL=https://api.example.com/openapi.json \
-e OPENAPI_TOKEN=your-token \
mcp-openapi-bridge
HTTP transport (for AWS Bedrock AgentCore or other hosted deployments):
docker build -f Dockerfile.agentcore -t mcp-openapi-bridge-agentcore .
docker run -p 8080:8080 \
-e API_BASE_URL=https://api.example.com \
-e OPENAPI_SPEC_URL=https://api.example.com/openapi.json \
-e OPENAPI_TOKEN=your-token \
-e MCP_AUTH_TOKEN=your-mcp-access-token \
mcp-openapi-bridge-agentcore
For public-routable deployments, set MCP_AUTH_TOKEN and configure clients to send Authorization: Bearer <token> to /mcp.
Spec caching
When using OPENAPI_SPEC_URL, the spec is cached locally for 1 hour (in ~/.cache/mcp-openapi-bridge/). Override with OPENAPI_SPEC_CACHE_TTL_SECONDS.
Limitations (v1)
- OpenAPI 3.x only (not Swagger 2.0)
- Internal
$refonly (no cross-file references) application/jsonbodies only (no multipart/form-data)- No cookie parameters
- No OAuth2 token flows (static token only)
- No per-operation auth override
allOf/anyOf/oneOfschemas fall back toz.any()
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.