Skip to content

flux7 docs

Governance infrastructure for AI agents. Three projects, one system.

agent7 (management plane)              ← visibility + control
  │              │
  │ reads via    │ reads via
  │ HTTP API     │ Python SDK
  │              │
  ▼              ▼
agent-mesh      mem7                   ← runtime + memory
(sidecar)       (memory substrate)
  │
  └──► tools (MCP, OpenAPI, CLI)

agent-mesh enforces what agents can do. Policy engine, approval queue, rate limiting, tracing. One Go binary, one YAML config.

mem7 remembers what happened. Persistent, searchable, governed memory. Hybrid search (BM25 + dense cosine + LLM reranking). Single Go binary.

agent7 shows it all and lets humans intervene. Dashboard, approval UI, governance scoring, audit trail. Next.js + FastAPI.

Progressive adoption

Step What you add What you get
Day 1 agent-mesh Policies, tracing, approval queue
Day 30 + mem7 Persistent memory, decision history, auto-approve
Day 60 + agent7 Dashboard, team approval UI, governance scoring

Each step is independently valuable. Start with agent-mesh — five minutes from install to first governed tool call.

Source