Social data for AI agents

Give agents live, public social data in a predictable shape. Connect any MCP client to the hosted MCP server, install the Agent Skill, or point a model at llms.txt.

Hosted MCP server

Nine tools over Streamable HTTP at mcp.crawlfeed.dev, with your API key as a header. Nothing to install.

Agent Skill

One command teaches your agent every endpoint, what it costs, and how to page and handle errors.

Predictable output

The same fields everywhere, so prompts and parsers do not branch per platform.

Budget-aware results

Every tool result states the credits it used and what is left, so an agent can stop before it overspends.

Connect your agent

Three ways in, one key and one credit balance behind all of them.

MCPDrop-in for any MCP client

Native tools for every MCP-compatible agent

One config block and your agent gets 9 tools: profiles, posts, comments, search across platforms, any of the endpoints in the catalogue, Arabic enrichment and your balance. Every result states what it cost, and errors carry a stable code and a retry hint.

  • Claude Code
  • Claude Desktop
  • Cursor
  • VS Code
  • Windsurf
  • Codex
mcp.json
{
  "mcpServers": {
    "crawlfeed": {
      "type": "http",
      "url": "https://mcp.crawlfeed.dev/mcp",
      "headers": {
        "Authorization": "Bearer sk_your_key_here"
      }
    }
  }
}
Setup for every client →
SkillClaude Code · Codex · Cursor

Install the skill in one command

Your agent learns every endpoint and what it costs, the response shape, paging, async search and error handling, so it calls the right thing the first time.

bash
npx skills add ilyasabdul19/crawlfeed-skills
llms.txtFor any model

Point any LLM at the docs

Every platform, endpoint and price in plain text, sized for a context window. The full reference is one file at llms-full.txt.

bash
curl -s https://crawlfeed.dev/llms.txt

Questions

Is there an MCP server?

Yes. Point any MCP client at https://mcp.crawlfeed.dev/mcp and send your key as "Authorization: Bearer sk_live_…". Clients that only run local servers can use npx mcp-remote. Setup snippets for each client are in the dashboard under Integrations.

Which tools does it expose?

get_profile, get_posts, get_post, get_comments, search (one platform or many), enrich_arabic, get_balance, list_endpoints (free, no key needed) and run_request, which reaches any endpoint in the catalogue.

Does the MCP server cost extra?

No. Tools spend the same credits as the equivalent API call, from the same balance, and failed calls are refunded.

How do I describe the API to my agent without MCP?

Install the skill with npx skills add ilyasabdul19/crawlfeed-skills, point it at crawlfeed.dev/llms.txt, or give it the OpenAPI spec at api.crawlfeed.dev/openapi.json.

Recipes for you