↑
onlyamicrowave 3 hours ago
I built AgentSift to solve a problem I kept hitting when building AI agents: every platform has its own capability registry, and there's no single place to search across all of them.
*What it does:* - Unified search across 7 agent platforms (MCP, LangChain, CrewAI, Composio, OpenAI, OpenClaw, NEAR AI) - 341+ indexed capabilities (tools, skills, integrations) - Fuzzy intent search (type "send email" and get every email-related capability, not just exact string matches) - Safety ratings: every capability shows permission levels and risk ratings
*Why the safety angle matters:* When you're giving an agent a "file management" capability, you need to know if that means "list filenames" or "recursively delete everything." AgentSift shows permission breakdowns for every capability.
*Tech:* - Next.js 15 + PostgreSQL - Fuse.js for fuzzy search - Platform-specific scrapers that run nightly to keep the index updated
*What's useful about this:* Instead of checking MCP's server registry, then LangChain's tool docs, then Composio's integration list… you search once. Cross-platform capability discovery in one place.
Still early, but it's live and usable. Planning to add a submission flow so developers can add capabilities directly, plus an API for agents to query programmatically.
Built this in about 6 hours during a weekend sprint. Feedback welcome.