OpenHands

OpenHands is a self-hosted developer control center that turns coding agents into an always-on engineering team. It runs the open-source OpenHands agent out of the box and can also drive any third-party agent such as Claude Code, Codex, or Gemini through the Agent-Client Protocol.

It runs locally by default but can connect to multiple agent backends, executing agents in Docker containers, on VMs, or across your own infrastructure. Beyond interactive conversations, it lets you build automations and workflows that integrate with Slack, GitHub, Linear, and other services, triggered on a schedule or by webhook events.

What You Get

• A self-hosted control center for coding agents that runs locally or on your own infrastructure
• The open-source OpenHands agent bundled and ready to use out of the box
• Support for third-party agents (Claude Code, Codex, Gemini) via the Agent-Client Protocol
• Pluggable agent backends across local machines, Docker containers, VMs, and cloud
• Automations and workflows that integrate with Slack, GitHub, Linear, and Notion on schedules or webhooks

Common Use Cases

• Running autonomous coding agents to implement features, fix bugs, and refactor code
• Automating repetitive engineering chores like report generation and issue decomposition
• Self-hosting agents inside company infrastructure for security and compliance
• Standardizing on one control plane across multiple agents and LLM providers

Under The Hood

Architecture - OpenHands is organized as a multi-part platform. The openhands/ Python package implements the agent runtime, event/action loop, and server (FastAPI), while a React frontend/ provides the Agent Canvas UI. Agents execute against pluggable backends that run in Docker containers, VMs, or cloud infrastructure, communicating via the Agent-Client Protocol so third-party agents can plug in. The core OpenHands agent and agent server are being split into the OpenHands/software-agent-sdk repository, with this repo hosting the control center.

Tech Stack - Python >=3.12 built on FastAPI, aiohttp, anyio, the Anthropic SDK, fastmcp, docker, gitpython, boto3, and browsergym-core, with a TypeScript/React frontend. Packaging uses Poetry with poetry-dynamic-versioning and uv, and there is an enterprise/ module under a separate license.

Code Quality - The repository is large and actively maintained, with a dedicated tests/ tree, pytest configuration, a Makefile, CI tooling, and dev_config for linting and formatting. Consistent release automation (release-please) and a strong contributor base back its maturity.

API Design - The product surface is primarily the Agent Canvas UI and CLI rather than a small importable API, so the developer experience centers on configuration: choosing a backend, selecting an agent, setting an LLM profile, and wiring automations. Sensible local defaults let you start immediately, while ACP support keeps agent integration standardized.