{"$schema":"https://wellknown.network/schemas/agent-record-v1.json","schemaVersion":"1","id":"ag_6z2ufv8xwhmz","handle":"agensflow-mcp","url":"https://wellknown.network/agents/agensflow-mcp","links":{"self":"https://wellknown.network/agents/agensflow-mcp/record.json","html":"https://wellknown.network/agents/agensflow-mcp","markdown":"https://wellknown.network/agents/agensflow-mcp/record.md","api":"https://wellknown.network/api/v1/agents/agensflow-mcp","status":"https://wellknown.network/api/v1/agents/agensflow-mcp/status","claim":"https://wellknown.network/agents/agensflow-mcp/claim","claimApi":"https://wellknown.network/api/v1/claims","claimDescriptor":"https://wellknown.network/agents/agensflow-mcp/claim.json","badge":"https://wellknown.network/agents/agensflow-mcp/badge.svg","openapi":"https://wellknown.network/openapi.json","history":"https://wellknown.network/api/v1/agents/agensflow-mcp/history","tools":"https://wellknown.network/api/v1/agents/agensflow-mcp/tools"},"ard":{"identifier":"urn:air::server:agensflow-mcp","type":"application/mcp-server-card+json"},"kind":"mcp_server","declared":{"name":"agensflow-mcp","summary":"AgensFlow MCP — learned coordination substrate for coding-agent loops","description":"# agensflow-mcp\n\nLearned coordination substrate for coding-agent and MAS loops. Apache-2.0.\nSelf-host, or use the hosted service.\n\n## What this is\n\nAgensFlow MCP is a policy server. For each task signature your agent sees, it\nlearns which sub-agent, model, and effort setting produces the best outcome —\nwhere \"best\" is a weighted trade-off of quality, cost, latency, and token\nbudget you configure.\n\nUnder the hood: a per-signature UCB1 bandit with Welford-online variance,\nAnthropic cache-aware pricing, tenant isolation, and BYOK judge orchestration.\n\nTwo integration paths cover essentially every agent shape:\n\n- **Framework adapters** — LangGraph, CrewAI, or any Runnable-shaped MAS\n  framework plug in via [`agensflow-langgraph`](https://github.com/AgensFlow-ai/agensflow-langgraph):\n  one decorator per node, per-node model pool, one line to record reward. The\n  adapter owns judge orchestration; the server records decisions and updates\n  the policy.\n- **Direct MCP integration** — Claude Code / codex hook the server via\n  PreToolUse + SubagentStop hooks. The server orchestrates a BYOK\n  single-model or 3-panel cross-family judge over `POST /judge/relative`.\n\n## Install\n\n```bash\npip install agensflow-mcp\n```\n\nRequires Python ≥ 3.11. Postgres is recommended for production; SQLite works\nout-of-the-box for local dev.\n\n## Run the server\n\n```bash\n# Local dev (in-process SQLite, no external deps)\nagensflow-mcp serve\n# → http://localhost:8000/health\n# → http://localhost:8000/docs      (OpenAPI schema explorer)\n# → http://localhost:8000/mcp/sse   (FastMCP endpoint)\n\n# Full stack with Postgres via Docker\ndocker run -d --name agensflow-postgres \\\n  -e POSTGRES_PASSWORD=agf -p 5432:5432 postgres:16\nAGF_DATABASE_URL=postgresql+psycopg://postgres:agf@localhost:5432/agensflow \\\n  agensflow-mcp serve\n```\n\nOnce up, verify:\n\n```bash\ncurl -s http://localhost:8000/health\n# → {\"status\":\"ok\",\"version\":\"...\"}\n```\n\n## Architecture\n\n```\n      Framework                Claude Code\n      adapter   …","publisher":{"name":"Nicole Koenigstein","url":null},"homepage":null,"repository":null,"version":"0.1.0","license":"Apache-2.0","protocols":["mcp"],"tags":["mcp"],"pricing":null,"endpoints":[{"url":"pypi:agensflow-mcp","type":"package_pypi","auth":null,"probeable":false}],"skills":null,"tools":null,"extra":null,"attribution":{"kind":"pypi","name":"pypi","license":"pypi","summary":"pypi","version":"pypi","description":"pypi","publisherName":"pypi"}},"derived":{"capabilities":[{"slug":"data.database","name":"Databases","confidence":1,"provenance":"derived"}],"categories":["data"],"language":"en"},"observed":{"status":"unknown","statusReason":"Distributed as a package to run locally; no network endpoint to check.","lastOkAt":null,"lastProbedAt":null,"statusComputedAt":null,"reliability30d":null,"latestObservations":[],"tools":null,"package":{"name":"agensflow-mcp","registry":"pypi","observedAt":"2026-11-08T22:24:46.209Z","publishedAt":"2026-07-18T09:04:48.280467Z","latestVersion":"0.1.0"},"toolSurface":null,"endpointFacts":[]},"verification":{"claimed":false,"claimedAt":null,"proofs":[]},"provenance":{"sources":[{"source":"pypi","key":"agensflow-mcp","url":"https://pypi.org/project/agensflow-mcp/","firstSeenAt":"2026-09-09T08:20:42.837Z","fetchedAt":"2026-11-08T22:23:32.214Z","normalizedAt":"2026-11-08T22:23:32.214Z"}]},"firstSeenAt":"2026-09-09T08:20:42.837Z","updatedAt":"2026-11-08T22:24:46.209Z"}