Smart Search: AI-Powered MCP Server Discovery
Finding the right MCP server among 23,000+ options shouldn't require guesswork. Smart Search uses semantic embeddings to match your intent, not just keywords.
The Discovery Problem
You need a tool that can send emails. Do you search for "email"? "SMTP"? "messaging"? "notification"? With traditional keyword search, you'd need to guess the exact terminology the server author used. Miss the right keyword and you miss the right tool.
This is the discovery problem. As the MCP ecosystem grows β AgentForge now hosts 106 servers with 400+ tools β keyword search becomes a bottleneck. AI agents need better discovery. So do developers.
How Smart Search Works
Smart Search uses semantic embeddings to understand what you mean, not just what you type. Here's the process:
- Every server description is converted into a high-dimensional vector using AI embeddings
- Your search query is embedded the same way, capturing intent and context
- Cosine similarity ranks servers by how closely they match your meaning
- Results surface even when zero keywords overlap with the query
Search for "I need to track customer payments" and you'll find Stripe, Square, and invoice tools β even if none of them contain the word "track" in their descriptions.
Real Examples That Show the Difference
Consider these searches and what Smart Search returns:
- "automate my deployment pipeline" finds GitHub Actions, Vercel, Docker, and CI/CD servers
- "talk to my database" surfaces Supabase, PostgreSQL, and SQLite MCP servers
- "help me write better code" matches linters, code review tools, and Playwright testing
Traditional search would fail on most of these. Semantic search understands the intent behind natural language.
Why It Matters for the Agent Economy
AI agents don't browse registries like humans do. They issue queries in natural language: "Find me a tool that can resize images" or "I need to check website uptime." If your search engine can only match keywords, agents will miss relevant tools β and tool providers will miss potential users.
Smart Search makes the AgentForge registry natively agent-friendly. Agents can describe what they need in plain language and get back the right MCP server every time. This is foundational infrastructure for the agent economy.
Built on Open Standards
Smart Search is powered by state-of-the-art embedding models and integrated directly into the AgentForge Discovery API. It works for both human developers browsing the registry and AI agents calling GET /api/v1/discover?q=your+query&smart=true.
No premium tier required β Smart Search is available to everyone. Because better discovery means a healthier ecosystem for all participants.
Try it now at /registry β toggle the Smart Search switch and search with natural language.