1. Click on "Install Server". 2. Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state. 3. In the chat, type @ followed by the MCP server name and your instructions, e.g., "@mcpify turn the Petstore API into an MCP server" That's it! The server will respond to your query, and you can continue using it as needed. Here is a step-by-step guide with screenshots.

mcpify

by bahri-hirfanoglu

Overview Schema Related Servers Score Discussions

TypeScript

Hybrid

Install

npm install -g @bahridev/mcpify

Quick Start

# Start an MCP server from a local spec
mcpify ./openapi.yaml

# From a URL
mcpify https://petstore3.swagger.io/api/v3/openapi.json

# With authentication
mcpify ./api.yaml --bearer-token $API_TOKEN

# HTTP transport (remote access)
mcpify ./api.yaml --transport http --port 3100

# Preview tools without starting server
mcpify ./api.yaml --dry-run

# Watch for spec changes
mcpify ./api.yaml --watch

# Interactive config setup
mcpify init

# Validate a spec for MCP compatibility
mcpify validate ./api.yaml

# Inspect a single tool
mcpify inspect ./api.yaml listPets

# Add to Claude Desktop config automatically
mcpify install ./api.yaml --name my-api --bearer-token $API_TOKEN

# Retry on 429/5xx, cache GETs, auto-paginate, pluck fields
mcpify ./api.yaml --retry 3 --cache-ttl 60 --auto-paginate \
  --response-fields "data[].id,data[].name"

# Compare two specs for breaking changes
mcpify diff ./old.yaml ./new.yaml --fail-on-breaking

# Smoke-test safe operations against a live API
mcpify test ./api.yaml --bearer-token $TOKEN

Usage

mcpify <spec> [options]

Arguments:
  spec                     OpenAPI spec file path or URL

Options:
  --spec <source>              Spec source (alternative to positional arg)
  --transport <type>           stdio (default) | http
  --port <number>              HTTP port (default: 3100)
  --base-url <url>             API base URL override
  --bearer-token <token>       Bearer token
  --api-key-header <name>      API key header name
  --api-key-value <value>      API key value
  --oauth-flow <flow>          OAuth2 flow (client_credentials | refresh_token)
  --oauth-token-url <url>      OAuth2 token endpoint
  --oauth-client-id <id>       OAuth2 client ID
  --oauth-client-secret <s>    OAuth2 client secret
  --oauth-refresh-token <t>    OAuth2 refresh token
  --oauth-scopes <scopes>      OAuth2 scopes (comma-separated)
  --include <patterns>         Include operations (glob, comma-separated)
  --exclude <patterns>         Exclude operations (glob, comma-separated)
  --tags <tags>                Only include these tags (comma-separated)
  --naming <style>             Tool naming: camelCase | snake_case | original
  --prefix <prefix>            Prefix for all tool names
  --header <key:value>         Custom headers (repeatable)
  --max-response-size <kb>     Max response size in KB (default: 50)
  --retry <count>              Retry attempts on 429/5xx (default: 0)
  --retry-delay <ms>           Base delay for retry backoff (default: 500)
  --retry-max-delay <ms>       Max delay for retry backoff (default: 10000)
  --cache-ttl <seconds>        Cache TTL for GET responses (default: 0)
  --cache-max <count>          Max cached entries (default: 100)
  --auto-paginate              Follow pagination links and merge pages
  --max-pages <count>          Max pages when paginating (default: 10)
  --response-fields <paths>    Dotted paths to select (e.g. "data[].id")
  --dry-run                    List tools without starting server
  --watch                      Watch spec file and reload on changes
  --verbose                    Verbose HTTP logging to stderr
  -V, --version                Output version
  -h, --help                   Show help

Commands:
  init                         Interactively create a .mcpifyrc.json
  validate <spec>              Report MCP compatibility of a spec
  inspect <spec> <tool>        Show full schema and example call for a tool
  install <spec>               Add entry to claude_desktop_config.json
  diff <left> <right>          Compare two specs, report changes
  test <spec>                  Smoke-test safe operations against a live API

Retry, Cache, Pagination, Field Selection

For resilient and token-efficient runtime behavior:

# Retry on 429/5xx with exponential backoff, honoring Retry-After
mcpify ./api.yaml --retry 3 --retry-delay 500 --retry-max-delay 10000

# Cache GET responses for 60 seconds (per-auth keyed)
mcpify ./api.yaml --cache-ttl 60 --cache-max 200

# Follow Link/next/cursor pagination and merge pages into one response
mcpify ./api.yaml --auto-paginate --max-pages 20

# Pluck only the fields the LLM needs from large responses
mcpify ./api.yaml --response-fields "data[].id,data[].name"

All four can be combined and are also configurable via .mcpifyrc.json (retry, retryDelay, retryMaxDelay, cacheTtl, cacheMax, autoPaginate, maxPages, responseFields).

`mcpify diff`

Compare two spec versions and report added, removed, and changed operations:

$ mcpify diff ./v1.yaml ./v2.yaml

Minimal API v0.1.0  →  Minimal API v0.2.0

Added:     1
Removed:   1
Changed:   1
Unchanged: 1

Added operations:
  + getStatus  GET /status

Removed operations:
  - legacy  GET /legacy

Changed operations:
  ~ getItem  GET /items/{id}
      + param query:verbose

Use --fail-on-breaking in CI to guard against removed operations, newly-required parameters, or method/path changes.

`mcpify test`

Smoke-test safe GET / HEAD operations against a live API:

$ mcpify test ./api.yaml --bearer-token $TOKEN

Base URL: https://api.example.com
Operations: 5  (ok: 3, fail: 0, skipped: 2)

  ✓ healthCheck  GET /health 200 45ms
  ✓ listPets     GET /pets   200 112ms
  ∘ getPet       GET /pets/{petId} (requires parameters)

Exits 1 when any probe fails. Operations requiring path/query/header parameters or using unsafe methods are skipped automatically.

Supported Specs

OpenAPI 3.0.x and 3.1.x
Swagger 2.0
YAML and JSON formats
Local files and remote URLs

Version is auto-detected — just point mcpify at any spec.

Claude Desktop Configuration

The fastest way: let mcpify edit the config for you.

mcpify install ./path/to/openapi.yaml --name my-api --bearer-token $API_TOKEN

This writes (or updates) an mcpServers.my-api entry in the platform-specific config file:

Platform	Path
Windows	`%APPDATA%\Claude\claude_desktop_config.json`
macOS	`~/Library/Application Support/Claude/claude_desktop_config.json`
Linux	`$XDG_CONFIG_HOME/Claude/claude_desktop_config.json` or `~/.config/Claude/claude_desktop_config.json`

Bearer tokens are written to entry.env (not to the CLI args) so they don't show up in process listings. Pass --force to overwrite an existing entry, --config <path> to target a different file, and --transport http --port 3100 to wire up the HTTP transport.

Or edit the file manually:

{
  "mcpServers": {
    "my-api": {
      "command": "mcpify",
      "args": ["./path/to/openapi.yaml"],
      "env": {
        "MCPIFY_BEARER_TOKEN": "your-token"
      }
    }
  }
}

HTTP Transport

Run as a remote MCP server accessible over HTTP:

mcpify ./api.yaml --transport http --port 3100

This exposes:

POST /mcp — MCP JSON-RPC endpoint (Streamable HTTP)
GET /mcp — SSE stream for server-initiated messages
DELETE /mcp — End an MCP session
GET /health — Health check endpoint

Each client gets an isolated session with its own MCP server instance. Sessions are tracked via the mcp-session-id header.

OAuth2

mcpify supports OAuth2 client credentials and refresh token flows. When either the oauth2 security scheme in your spec exposes a clientCredentials.tokenUrl / authorizationCode.refreshUrl and the corresponding credentials are present in the environment, mcpify will automatically fetch tokens and refresh them before expiry.

Client credentials (machine-to-machine)

mcpify ./api.yaml \
  --oauth-flow client_credentials \
  --oauth-token-url https://auth.example.com/token \
  --oauth-client-id $CLIENT_ID \
  --oauth-client-secret $CLIENT_SECRET \
  --oauth-scopes "read,write"

Or equivalently, let mcpify discover tokenUrl from the spec and only supply the credentials via env vars:

export MCPIFY_OAUTH_CLIENT_ID=...
export MCPIFY_OAUTH_CLIENT_SECRET=...
mcpify ./api.yaml

Refresh token

mcpify ./api.yaml \
  --oauth-flow refresh_token \
  --oauth-token-url https://auth.example.com/token \
  --oauth-client-id $CLIENT_ID \
  --oauth-refresh-token $REFRESH_TOKEN

The refresh token is rotated automatically when the server returns a new one in the token response.

`.mcpifyrc.json`

{
  "spec": "./api.yaml",
  "oauth": {
    "flow": "client_credentials",
    "tokenUrl": "https://auth.example.com/token",
    "clientId": "my-client",
    "clientSecret": "shhh",
    "scopes": ["read", "write"]
  }
}

Authorization code / OpenID Connect

Full interactive browser-based authorization is out of scope for now. Use your IdP's tooling (or a one-time curl exchange) to obtain a refresh token, then pass it with --oauth-refresh-token. If your spec uses an openIdConnect scheme, provide --oauth-token-url explicitly because mcpify does not auto-discover OIDC configuration.

`mcpify init`

Generates a .mcpifyrc.json interactively:

$ mcpify init
OpenAPI spec source (file path or URL): ./openapi.yaml
Transport (stdio | http) (stdio): http
HTTP port (3100):
Override base URL (empty to skip):
Auth type (none | bearer | api-key | oauth2) (none): oauth2
OAuth2 flow (client_credentials | refresh_token) (client_credentials):
...
✓ Wrote /path/to/.mcpifyrc.json

Pass --force to overwrite an existing config without confirmation.

`mcpify validate`

Reports which parts of a spec mcpify can handle, with issues grouped by severity:

$ mcpify validate ./api.yaml

OAuth API v1.0.0
Base URL: https://api.example.com

Operations: 12
Tools:      12

Security schemes:
  ✓ oauth (oauth2 (clientCredentials))
  ⚠ oidc (openIdConnect)

Issues:
  ⚠ security scheme "oidc" is OpenID Connect — mcpify does not
    auto-discover the tokenUrl. Provide --oauth-token-url manually

⚠ PASS with warnings — 1 warning(s)

Exits with code 1 when errors are present. Use this in CI to guard against spec regressions.

`mcpify inspect`

Shows the full tool metadata, a placeholder example argument object, and an equivalent cURL invocation:

$ mcpify inspect ./api.yaml getPet

getPet
──────
GET /pets/{petId}
Base URL: https://petstore.example.com/v1
Tags: pets
Hints: read-only

Description:
  Get a pet by ID

  Returns: {id, name, tag}

Input schema:
  { ... }

Example arguments:
  { "petId": "<string>" }

cURL:
  curl -X GET 'https://petstore.example.com/v1/pets/%3Cstring%3E' \
    -H 'Accept: application/json'

Respects --naming, --prefix, --include, --exclude, --tags, and --base-url, so tool name resolution matches the configuration you use at runtime.

Docker

# Build
docker build -t mcpify .

# Run over HTTP
docker run --rm -p 3100:3100 \
  -v "$PWD/openapi.yaml:/spec/openapi.yaml:ro" \
  mcpify /spec/openapi.yaml --transport http --port 3100

Published images are available at ghcr.io/bahri-hirfanoglu/mcpify:<version>.

GitHub Action

Use the composite action to validate specs or spin up an MCP server in a workflow:

- uses: bahri-hirfanoglu/mcpify@v1
  with:
    spec: ./openapi.yaml
    command: validate

Supported commands: validate, dry-run, inspect, serve. Pass additional flags via extra-args.

Config File

Create a .mcpifyrc.json in your project root to avoid repeating CLI flags:

{
  "spec": "./openapi.yaml",
  "transport": "stdio",
  "bearerToken": "sk-...",
  "include": ["get*", "list*"],
  "exclude": ["delete*"],
  "naming": "snake_case",
  "prefix": "myapi_",
  "headers": {
    "X-Custom-Header": "value",
    "X-Api-Version": "2024-01"
  },
  "verbose": true
}

Supported file names: .mcpifyrc.json, .mcpifyrc, mcpify.config.json

CLI flags always override config file values.

Custom Tool Naming

# Convert operationIds to snake_case
mcpify api.yaml --naming snake_case

# Add a prefix to all tool names
mcpify api.yaml --prefix myapi_

# Combine both
mcpify api.yaml --naming snake_case --prefix myapi_
# listAllPets → myapi_list_all_pets

Custom Headers

Send custom HTTP headers with every API request:

# Single header
mcpify api.yaml --header "Authorization: Bearer sk-..."

# Multiple headers
mcpify api.yaml --header "Authorization: Bearer sk-..." --header "X-Api-Version: 2024-01"

Also configurable via .mcpifyrc.json (headers field).

Filtering Operations

# Only include specific operations
mcpify api.yaml --include "get*,list*"

# Exclude destructive operations
mcpify api.yaml --exclude "delete*,remove*"

# Filter by tags
mcpify api.yaml --tags "users,pets"

Response Schema Hints

Tool descriptions automatically include response structure information extracted from the spec:

listPets — List all pets

Returns: array of {id, name, tag}

This helps AI assistants understand what the API returns before calling a tool.

Key Sanitization

mcpify automatically sanitizes JSON Schema property keys that contain special characters (dots, brackets, hyphens, etc.) to ensure compatibility with all MCP clients. Original keys are restored when making API requests, so the target API always receives the correct parameter names.

X-Request-ID  →  x_request_id  (in tool schema)
filter[name]  →  filter_name_  (in tool schema)

This is fully transparent — no configuration needed.

Verbose Logging

mcpify api.yaml --verbose

Logs HTTP requests and responses to stderr:

→ GET https://api.example.com/pets?limit=10
← ✓ 200 45ms 1.2KB
→ POST https://api.example.com/pets
← ✗ 401 12ms 156B

Environment Variables

Variable	Description
`MCPIFY_BEARER_TOKEN`	Bearer token for authentication
`MCPIFY_API_KEY_HEADER`	API key header name
`MCPIFY_API_KEY_VALUE`	API key value
`MCPIFY_OAUTH_CLIENT_ID`	OAuth2 client ID
`MCPIFY_OAUTH_CLIENT_SECRET`	OAuth2 client secret
`MCPIFY_OAUTH_REFRESH_TOKEN`	OAuth2 refresh token

Programmatic API

import { parseSpec, generateTools, startServer } from '@bahridev/mcpify';

const spec = await parseSpec('./openapi.yaml');
const tools = generateTools(spec.operations);

await startServer({
  spec,
  tools,
  operations: spec.operations,
  baseUrl: spec.defaultServerUrl,
  auth: { type: 'none' },
  transport: 'stdio',
  port: 3100,
  maxResponseSize: 50 * 1024,
});

How It Works

Parses and dereferences the OpenAPI/Swagger spec
Converts each operation to an MCP tool with JSON Schema input
Adds response schema hints to tool descriptions
Starts an MCP server (stdio or HTTP transport)
When a tool is called, builds and executes the corresponding HTTP request
Returns the API response as MCP tool output

License

MIT

This server cannot be installed

license - permissive license

quality - not tested

maintenance

How are these scores calculated?

Maintenance

–Maintainers

–Response time

1dRelease cycle

5Releases (12mo)

Resources

Need Help?

Related Servers

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

Latest Blog Posts

Lightport: Open-Sourcing Glama's AI Gateway
By punkpeye on April 27, 2026.
open source
OpenAI
Tool Definition Quality Score (TDQS)
By punkpeye on April 3, 2026.
mcp
The Hackers Who Tracked My Sleep Cycle
By punkpeye on March 26, 2026.
security

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/bahri-hirfanoglu/mcpify'

If you have feedback or need assistance with the MCP directory API, please join our Discord server