Integrations

Agent Skills

Install the Social Fetch agent skill with npx skills — teach Cursor, Claude, and other agents to use llms.txt, MCP, POST /v1/ask, and real endpoints only.

Give coding agents a short onboarding pack for Social Fetch: how to read /llms.txt and /agents.txt, use per-platform /llms-{platform}.txt files, connect MCP, call POST /v1/ask, create an API key, and stop inventing routes that do not exist.

Install

npx skills add social-freak-ltd/socialfetch -s socialfetch

Useful flags: -g for a user-level install, -y to skip prompts, -a cursor (or claude-code, etc.) to target one agent.

Install from the skill path only (avoids other skills that may live in this monorepo):

npx skills add https://github.com/social-freak-ltd/socialfetch/tree/main/skills/socialfetch

Browse the ecosystem at skills.sh. Source: skills/socialfetch in the Social Fetch repo.

What the skill teaches

  • Read /llms.txt and /agents.txt, then /llms-endpoints.txt or /llms-{platform}.txt, before writing integration code
  • Connect hosted MCP at https://api.socialfetch.dev/mcp with OAuth (no API key in MCP config)
  • Use POST /v1/ask / MCP nl_ask_post to explore, then pin the typed route
  • Create a sfk_… key at API Keys for REST and the SDK
  • Treat lookupStatus and meta.requestId correctly — never invent endpoints

After install, ask your agent to fetch Social Fetch data or scaffold a lookup. It should load the skill when Social Fetch is in scope.

Skills vs MCP vs llms.txt

PieceRole
Agent SkillDurable instructions in the agent harness (when / how to use Social Fetch)
MCPLive tools — call endpoints and docs search while coding
/llms.txt / /agents.txtPublic orientation and crawl/onboarding files
/llms-{platform}.txtPer-platform parameters, credits, and example curls

You can use any combination. Skill + MCP is the usual Cursor/Claude setup; REST + skill works for custom orchestrators.

On this page