[Feature Request] Integrate MCP Discovery for dynamic tool discovery

[yksanjo]

Hi AutoGPT team,

I'd like to propose integrating MCP Discovery API to enable dynamic tool discovery for AutoGPT agents.

Problem

Agents are limited to pre-configured tools. There's no way to discover new capabilities dynamically based on the task.

Solution

MCP Discovery provides semantic search across MCP servers (Model Context Protocol), letting agents find and use the right tool for any task.

How it works

# Agent needs database capability
result = discover_mcp_server(need="database with authentication")
# Returns: Supabase MCP (94% match) + install command + uptime: 99.9%

Features

• Semantic search - Find tools using natural language
• Performance metrics - Latency, uptime, success rates
• 24+ servers indexed - Databases, communication, automation, file systems
• Free tier - 100 queries/month

Benefits for AutoGPT

• Agents can adapt their toolset based on task requirements
• No need to pre-configure every possible tool
• Performance data helps choose reliable options

Resources

• GitHub: https://github.com/yksanjo/mcp-discovery
• API: https://mcp-discovery-production.up.railway.app
• npm: https://www.npmjs.com/package/mcp-discovery-api

Happy to help with integration!