hive

Multi-Agent Harness for Production AI

⭐ 10,227 stars on GitHub · 🍴 5,651 forks · 📜 License: apache-2.0 · 💻 Language: Python

What is hive?

If your AI agents are getting past demos and into real workflows, you need more than prompt chains and optimistic retries. Hive positions itself as a production harness for multi-agent systems, with the strongest focus on state, recovery, observability, and human control rather than just agent abstraction.

Main components

• Multi-agent coordination for running specialized agents in parallel across complex workflows
• Graph-based execution DAGs that turn objectives into structured, repeatable processes
• Persistent role-based memory that carries project context across long-running jobs
• Failure recovery and self-healing behavior for agents that need to survive crashes and bad intermediate steps
• Model-agnostic runtime with support for OpenAI, Anthropic, Google Gemini, and custom models
• Human-in-the-loop controls, auditability, cost limits, and observability for production operations
• Browser-use and general compute capabilities for agents that need to interact with real systems

Clear use cases

• Automate multi-step business operations where several agents need to research, decide, execute, and verify work
• Run long-lived AI workflows that require persistent state, crash recovery, and resumable execution
• Build internal agent platforms where teams can plug in different LLM providers without rewriting orchestration logic
• Coordinate browser-based automation tasks with oversight, logging, and rollback-friendly execution
• Prototype production agent workflows before committing to a commercial orchestration platform

The biggest strength is production-minded orchestration — Hive is aimed at the messy runtime layer that many agent frameworks underplay: state persistence, fault handling, observability, cost control, and human approval. Compared with commercial agent platforms, its value is that you can self-host the harness, keep control of infrastructure and data, and still get a structured execution model for serious workflows instead of stitching together scripts, queues, and ad hoc retry logic.

Best for AI platform engineers, automation teams, and technical founders building multi-agent workflows that need to run reliably beyond a notebook or demo environment.

Topics: the project is tagged with popular topics:

• 🏷️ agent
• 🏷️ agent-framework
• 🏷️ agent-skills
• 🏷️ anthropic
• 🏷️ automation
• 🏷️ autonomous-agents
• 🏷️ claude
• 🏷️ harness
• 🏷️ harness-engineering
• 🏷️ human-in-the-loop

📸 Screenshots

Quick install

See the README for detailed install instructions. Most projects support Docker — if the repo has a Dockerfile, use:

git clone https://github.com/aden-hive/hive.git
cd hive
docker build -t hive .
docker run -d -p 8080:8080 hive

Minimum system requirements

⚡ Deploy fast on VSIS

Use the VSIS VPS Mini 1GB RAM / 1 vCPU / 15GB SSD (~70k/tháng) plan from VSIS.NET — high-speed VN-based VPS, 24/7 support, ideal for running hive smoothly.

🎯 Benefits:

• One-command docker compose up -d deploy in 2 minutes
• Dedicated IPv4, root access, unmetered domestic bandwidth
• Daily snapshot backup
• Free install assistance from the VSIS team

👉 See matching VPS plans at vsis.net

Resources

• 🔗 GitHub: aden-hive/hive
• 📚 Official docs: see README in the repo
• 💬 Community: GitHub Issues + Discussions

Article compiled from GitHub data on 05/05/2026. Star/fork counts may have changed — see live numbers via the GitHub link.