HiveFlow Documentation¶
HiveFlow is the open-source multi-agent coordination & HITL layer with a visual Studio and first-class MCP tool integration.
Status: 0.1.x Alpha — API may evolve; see Versioning.
Choose your path¶
| I want to… | Start here |
|---|---|
| Embed orchestration in Python | Core module · Getting Started |
| NL planning + Agent runtime | Agent module · Studio Agent cookbook |
| Visual workflows + ops UI | Studio module · Agent Operations |
| Migrate from LangGraph | Migration guide · LangGraph Sidecar |
Three modules¶
| Module | Description | Details |
|---|---|---|
Core (hiveflow) |
Scheduler, blackboard, DAG/HITL/RAG/MCP kernel | Modules overview |
Agent (hiveflow-agent) |
NL TaskGraph planning + cognitive execution | Modules overview |
| Studio | Self-hosted FastAPI + React ops UI | Modules overview |
Quick Links¶
- Getting Started — install and first workflow
- Core Concepts — ECM, Cell, Blackboard, Scheduler
- Cookbook — HITL, RAG, Studio Agent mode
- API Reference — auto-generated core API
- Architecture — three-layer design
- Deployment — Docker, Kubernetes
- Observability — metrics, tracing, Studio analytics
Minimal Example (Core)¶
import asyncio
from hiveflow import HiveFlow, HiveFlowConfig, ECM
async def main():
hf = HiveFlow(HiveFlowConfig())
await hf.start()
try:
async def handler(ecm, view):
await view.put("result", f"Done: {ecm.intent}")
await hf.create_agent("worker", {"process"}, set(), {"result"}, handler)
await hf.scheduler.schedule(ECM(
trace_id="1", intent="Hello", intent_id="1",
emitter="user", required_skills=["process"],
))
finally:
await hf.shutdown()
asyncio.run(main())
Community¶
- GitHub Repository
- Issue Tracker
- Discussions (enable in repo settings)
- Contributing
- Governance