api-mind
MCP server for API discovery from .mind spec files, enabling AI assistants to list, inspect, and call APIs defined in OpenAPI specs.
README
api-mind MCP Server
MCP server for API discovery from .mind spec files. Works with Claude Code, Claude Desktop, and any MCP-compatible AI assistant.
Quick Start
# 1. Install spec-mind to generate .mind files from OpenAPI specs
brew install spec-mind
# 2. Run the setup script from your project directory
curl -fsSL https://raw.githubusercontent.com/msegoviadev/api-mind-mcp/main/setup-mcp.sh | bash
# 3. Add your OpenAPI specs and generate .mind files
mkdir specs
cp your-api.yaml specs/
spec-mind sync --no-notation ./specs/
# 4. Restart Claude Code
In Claude Code:
What APIs are available?
Show me the endpoints for payments
Call GET /payments/{id} in dev
Installation
Prerequisites
Install spec-mind to generate .mind files from your OpenAPI specs:
brew install spec-mind
Setup
Run setup-mcp.sh from your project directory:
curl -fsSL https://raw.githubusercontent.com/msegoviadev/api-mind-mcp/main/setup-mcp.sh | bash
Or for user-wide installation (all projects):
curl -fsSL https://raw.githubusercontent.com/msegoviadev/api-mind-mcp/main/setup-mcp.sh | bash -s -- --global
The script:
- Creates a
specs/directory in your project - Registers api-mind in Claude Code with the correct specs path
- Scaffolds
~/.config/api-mind/dev.envfor environment defaults
Usage
1. Add your API specs
# Copy your OpenAPI/Swagger YAML/JSON files to specs/
cp your-api.yaml specs/
2. Generate .mind files
spec-mind sync --no-notation ./specs/
3. Use in Claude Code
User: "What APIs are available?"
Claude: *uses list_apis tool*
"Found 2 APIs: ecommerce, payments"
User: "Show me payment endpoints"
Claude: *uses list_endpoints tool*
"POST /payments [auth: oauth2]
GET /payments/{id}"
User: "Call GET /payments/{id} in dev"
Claude: *uses get_endpoint_schema + get_call_context tools*
"Resolved base URL: https://api.dev.example.com
Calling GET /payments/123..."
Tools
list_apis
Lists all APIs loaded from the specs folder.
Input: none
Output: JSON with API names, titles, base URLs, and environments
list_endpoints
Lists endpoints across all APIs.
Input:
filter (optional): Substring match on method, path, or section
Output: JSON with environments and endpoint list
get_endpoint_schema
Returns full context for an endpoint.
Input:
api: API name
method: HTTP method
path: Endpoint path
Output: Text block with base URL, environments, auth, and schema
Call before constructing curl to understand the endpoint contract.
get_call_context
Returns runtime context needed to execute API calls.
Input:
api: API name
env (optional): Environment to use (dev, stage, uat). Defaults to dev.
Output: Resolved base URL, active environment, and default values for credentials and parameters
Call this before constructing curl when the user wants to actually invoke an endpoint.
Reads from ~/.config/api-mind/<env>.env and ~/.config/api-mind/<api>/<env>.env.
Auth Patterns
When get_endpoint_schema shows auth requirements, construct headers:
| Auth in Schema | curl Header |
|---|---|
None |
No header |
bearer |
-H 'Authorization: Bearer <TOKEN>' |
oauth2 <scopes> |
-H 'Authorization: Bearer <TOKEN>' |
api_key <header> |
-H '<header>: <KEY>' |
basic |
-H 'Authorization: Basic <base64>' |
Environment Defaults
get_call_context reads default values from ~/.config/api-mind/:
~/.config/api-mind/
dev.env # base defaults for all APIs (dev environment)
stage.env # base defaults for stage
auth0/
dev.env # API-specific overrides for auth0
Each .env file uses key=value format (lines starting with # are ignored). The base_url key overrides the placeholder URL from the spec.
Example ~/.config/api-mind/dev.env:
base_url=https://api.dev.example.com
auth0_client_id=abc123
auth0_cacert=/etc/ssl/cert.pem
setup-mcp.sh scaffolds this file on first run.
Workflow
list_apis → list_endpoints → get_endpoint_schema → get_call_context → [LLM constructs curl] → bash
list_apis- Discover available APIslist_endpoints- Find relevant endpointsget_endpoint_schema- Get endpoint contract (URL, auth, schema)get_call_context- Resolve base URL and credentials for the target environment- LLM constructs curl command using resolved values
- LLM executes via
bashtool
Advanced Configuration
<details> <summary>Manual installation and team setup</summary>
Manual Installation
claude mcp add --transport stdio api-mind \
--env SPECS_DIR=/absolute/path/to/specs \
-- npx -y @msegoviadev/api-mind-mcp
Important: Always use absolute paths. MCP servers run as standalone processes without project context.
Project Configuration (.mcp.json)
Create .mcp.json in your project root:
{
"mcpServers": {
"api-mind": {
"command": "npx",
"args": ["-y", "@msegoviadev/api-mind-mcp"],
"env": {
"SPECS_DIR": "/absolute/path/to/specs"
}
}
}
}
Note: Each team member needs their own .mcp.json with their absolute path. Add .mcp.json to .gitignore.
</details>
Development (Contributors)
git clone https://github.com/msegoviadev/api-mind-mcp
cd api-mind-mcp
npm install
npm run build
# Test locally
node dist/index.js /path/to/specs
Related
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.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
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.