| name | find-mcp-tools |
| description | Find MCP servers and MCP-capable tools with Can Agents Use. Use when a user asks for Model Context Protocol tools, MCP servers, remote MCP endpoints, local MCP setup, agent tool calls, or software that integrates directly with MCP-aware agents. |
| allowed-tools | WebFetch |
Find MCP Tools
Use this skill when the user wants tools that expose an MCP server or work well through MCP-aware agents such as Cursor, Claude Code, Codex, OpenCode, Gemini CLI, or similar tools.
Search
Prefer the capability filter:
curl -fsS 'https://canagentsuse.com/api/agent/search?capability=mcp&page=1&limit=20'
Search a specific domain plus MCP:
curl -fsS 'https://canagentsuse.com/api/agent/search?q=<domain>%20mcp&capability=mcp&page=1&limit=10'
Fetch details before recommending:
curl -fsS 'https://canagentsuse.com/api/agent/tools/<slug>'
MCP Endpoint For Can Agents Use
Agents can query Can Agents Use itself through MCP:
{
"mcpServers": {
"canagentsuse": {
"type": "http",
"url": "https://canagentsuse.com/api/mcp"
}
}
}
Useful MCP tools: search_agent_tools, get_agent_tool, get_agent_catalog, list_agent_capabilities, and get_agent_score_model.
Evaluate MCP Results
- Prefer
native MCP support over wrappers when the user needs reliability.
- Check whether the MCP server is remote, local, official, community-maintained, authenticated, or read-only.
- Look for sandbox or test-mode support before live actions.
- Include any setup command, package name, endpoint URL, auth notes, and caution notes.
- Do not treat MCP presence as approval to run destructive actions.