Exécutez n'importe quel Skill dans Manus
en un clic

Exécutez n'importe quel Skill dans Manus en un clic

developers.md

Developer Portal

Agent-Ready API

# Developer Portal

Integrate with the SkillsMP API to discover 1.2M+ Claude Code skills programmatically. Build agent-ready applications with our REST API, MCP server, and OpenAPI specification.

Full API Docs OpenAPI Spec

Keyword Search

Search 1.2M+ skills by keyword with filtering by category, occupation, and sort order.

AI Semantic Search

Natural language search using vector embeddings to find the most relevant skills.

MCP Server

Connect Claude, ChatGPT, and other AI agents directly to SkillsMP via Model Context Protocol.

Secure & Rate-Limited

API key authentication with daily quotas, burst protection, and comprehensive logging.

OpenAPI Spec

Auto-generated OpenAPI 3.0 specification for automatic client generation and discovery.

Multi-Language SDKs

cURL, JavaScript, Python examples. Generate typed clients from our OpenAPI spec.

Agent Authentication

Step-by-step guide for AI agents to authenticate programmatically with long-lived API keys.

Webhooks & Polling

How agents stay in sync with the catalog: polling patterns, timeline feed, and webhook roadmap.

quickstart.md

Quickstart Guide

Get an API Key

Make your first request

Search skills using the REST API. Replace YOUR_API_KEY with your actual key, or omit the header for anonymous access.

bash

curl -X GET "https://skillsmp.com/api/v1/skills/search?q=SEO" \
  -H "Authorization: Bearer YOUR_API_KEY"

javascript

const response = await fetch(
  'https://skillsmp.com/api/v1/skills/search?q=SEO',
  {
    headers: {
      'Authorization': 'Bearer YOUR_API_KEY'
    }
  }
);
const data = await response.json();
console.log(data.data.skills);

python

import requests

response = requests.get(
    'https://skillsmp.com/api/v1/skills/search',
    params={'q': 'SEO'},
    headers={'Authorization': 'Bearer YOUR_API_KEY'}
)

data = response.json()
print(data['data']['skills'])

Explore the API

Read the full API documentation for detailed endpoint references, error codes, and rate limiting information.

Read API Documentation

mcp.md

MCP Server

SkillsMP exposes a Model Context Protocol (MCP) server, allowing AI agents like Claude and ChatGPT to search and discover skills natively. Use Streamable HTTP transport for seamless integration.

Server Endpoint

POST https://skillsmp.com/mcp

Discovery

GET https://skillsmp.com/.well-known/mcp/server-card.json

Available Tools

search_skills

Search skills by keyword with filters

get_skill

Get detailed information about a specific skill

list_categories

Browse all skill categories and domains

ai_search_skills

Semantic search using natural language

sandbox.md

API Sandbox

Test the API directly in your browser. No API key required for basic search (anonymous quota: 50 requests/day).

GET /api/v1/skills/search

Try: search for "web scraping"

GET /api/health

Check API status

agent-auth.md

AI Agent Authentication

Step-by-step guide for AI agents (Claude, ChatGPT, autonomous pipelines, bots) to authenticate programmatically with the SkillsMP API. No OAuth or user interaction required at runtime — agents use long-lived API keys.

Provision an API key (one-time, human-in-the-loop)

The owning human signs in at /auth/login (Google or GitHub), opens the Developer Portal dashboard, and generates an API key. Store the key in the agent's secret manager (env var, vault, KMS).

Send the key on every request

Add the Authorization header. Treat the key as a bearer token — do not put it in the URL, query string, or client-side code.

http

Authorization: Bearer sk_live_your_api_key

Handle rate limits gracefully

Every response includes X-RateLimit-Daily-Remaining and X-RateLimit-Minute-Remaining. On 429, read the header, back off, and retry. Anonymous: 50/day · 10/min. Authenticated: 500/day · 30/min.

python

# Python: back off when remaining <= 0
r = requests.get(url, headers={'Authorization': f'Bearer {API_KEY}'})
if r.status_code == 429:
    time.sleep(60)  # wait for the minute window to reset
    r = requests.get(url, headers={'Authorization': f'Bearer {API_KEY}'})

Identify your agent (recommended)

Set a descriptive User-Agent so we can reach you if we detect issues. Example: SkillsMPClient/1.0 (+https://your-agent.example)

Prefer MCP for tool-calling agents

If your agent supports Model Context Protocol, connect to https://skillsmp.com/mcp instead of hand-rolling REST calls. MCP handles tool discovery, schema, and tool-use loops natively.

Security: API keys do not expire. Rotate from the Developer Portal if exposed. Anonymous requests (no key) work for keyword search but are heavily rate-limited and not recommended for production agents.

webhooks.md

Webhooks

SkillsMP does not currently emit outbound webhooks for new-skill or update events. Agents that need to track the catalog should poll the endpoints below — the ETL pipeline refreshes hourly, so a 1-hour cadence is sufficient for most use cases.

Poll for new/updated skills

Use sortBy=recent and compare the updatedAt timestamp to your last-seen cursor.

http

GET /api/v1/skills/search?q=*&sortBy=recent&limit=50

Subscribe via MCP (push-style)

MCP-enabled agents can call search_skills on a timer and stream results into their context — no outbound HTTPS endpoint required on your side.

Timeline feed

For aggregate growth/update trends, use the public timeline feed:

http

GET /api/timeline?granularity=daily&limit=30

Roadmap: Outbound webhooks for new-skill and update events are planned. If you have a concrete use case, open an issue on GitHub or reach us via the About page so we can prioritize.

policies.md

Rate Limits & Policies

Anonymous Access

Daily quota: 50 requests
Burst limit: 10 requests/minute (search only)
No API key required

Authenticated Access

Daily quota: 500 requests
Burst limit: 30 requests/minute (search only)
API key required in Authorization header