Agent Store MCP Server

Connect any MCP-compatible AI system to the Agent Store. Search, discover, and route tasks to the best AI agent — programmatically.

Server

agent-store

Version

1.0.0

Protocol

2025-11-25

Transport

HTTP

Operational

/.well-known/mcp.json /api/mcp

Quick Setup

Add the Agent Store as an MCP server in your client config.

claude_desktop_config.json

{
  "mcpServers": {
    "agent-store": {
      "url": "https://agentstore.dev/api/mcp"
    }
  }
}

Restart Claude Desktop after saving. The 4 agent discovery tools will appear automatically.

Capabilities

What this server exposes to connected clients.

Tools

4 tools for searching, filtering, and discovering agents by task

Resources

Full catalog and category directory as readable resources

Prompts

Coming soon

Sampling

Coming soon

Tools

Complete reference for all available tools.

find_agent_for_task

Describe a task in natural language and get ranked agent recommendations. This is the primary discovery endpoint — the routing layer for AI-to-AI delegation. Returns agents scored by relevance with full metadata.

Parameter	Type	Description
`task_description`required	string	Natural language description of the task
`max_results`	number	Max agents to return (default 5, max 20)
`pricing_preference`	string	"free" \| "freemium" \| "paid" \| "any"

search_agents

Structured search across the entire agent catalog. Supports text queries combined with category, capability, tool, and pricing filters. Returns paginated results.

Parameter	Type	Description
`query`	string	Free-text search
`category`	string	Category slug filter
`capability`	string	Filter by capability
`tool`	string	Filter by tool the agent uses
`pricing`	string	"free" \| "freemium" \| "paid" \| "contact"
`limit`	number	Max results (default 10, max 50)

get_agent

Get complete details for a specific agent by slug. Returns capabilities, tools, pricing, ratings, special data, and related agents in the same category.

Parameter	Type	Description
`slug`required	string	Agent identifier (e.g. "cursor", "perplexity")

list_categories

List all agent categories with descriptions and live agent counts. No parameters required.

Resources

Readable resources exposed by this server.

agentstore://catalog

Complete catalog of all approved agents with full metadata

application/json

agentstore://categories

All categories with descriptions and agent counts

application/json

Example: Find an agent for a task

A user asks Claude: “Find me an AI that can help with legal contract review.” Claude calls the Agent Store:

Request

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "find_agent_for_task",
    "arguments": {
      "task_description": "review and analyze legal contracts",
      "max_results": 2
    }
  }
}

Response

{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "content": [{
      "type": "text",
      "text": {
        "task": "review and analyze legal contracts",
        "recommendations": [
          {
            "name": "Harvey",
            "slug": "harvey",
            "tagline": "AI legal assistant for law firms",
            "capabilities": ["Contract drafting", "Document review", "Legal research"],
            "website_url": "https://harvey.ai",
            "pricing_type": "contact",
            "rating": 4.5,
            "relevance_score": 72
          }
        ],
        "total_matches": 1
      }
    }]
  }
}

Claude reads the response and recommends Harvey to the user, with pricing, rating, and a direct link.

REST API

Prefer REST over JSON-RPC? The same data is available via standard HTTP endpoints.

GET/api/agentsSearch & list agents GET/api/agents/:slugGet agent by slug GET/api/categoriesList all categories

List your agent on the store

Every listed agent is automatically discoverable via MCP by every connected AI system.

Submit Agent