Skip to content
RepoQuarry

OpenHands

Compare

Self-hosted control center and CLI for AI coding agents, running OpenHands or Claude Code and Codex with any model, including local LLMs.

Open-source alternative toClaude CodeCursor

OpenHands at a glance

Checked 21 Oct 2026

The OpenHands repo now holds Agent Canvas (v1.29.0, October 2026), a self-hosted control center that runs the open-source OpenHands agent, or Claude Code, Codex and other ACP agents, on your machine, in Docker, on a VM or in OpenHands Cloud. It adds automations such as issue-to-pull-request and PR review. A separate OpenHands CLI has a headless mode for CI and connects to VS Code, JetBrains and Zed over ACP. Bring any model, including local ones via LM Studio, Ollama, vLLM or SGLang. MIT licence; OpenHands Cloud is the optional paid service.

Cite this pageRepoQuarry. “OpenHands.” Checked 21 Oct 2026. https://repoquarry.dev/projects/openhands

How much of Claude Code does it replace?

Use it instead of paying for Claude Code?
Personal use
Confirmed: Yes
For a business
Confirmed: Yes

6 of 6 jobs. CLI, headless mode, MCP and issue-to-PR automations; VS Code support is through a community ACP extension, and results depend on the model you pick.

Partial: Edit code across a repo from the terminal with an AI agent Partial replacement
Partial: Run commands and tests Partial replacement
Partial: Use the agent inside VS Code or JetBrains Partial replacement
Partial: Connect tools and data through MCP Partial replacement
Partial: Write commits and pull requests Partial replacement
Partial: Run the agent headless in scripts or CI (GitHub Actions) Partial replacement

How much of Cursor does it replace?

Use it instead of paying for Cursor?
Personal use
Confirmed: Yes
For a business
Confirmed: Yes

4 of 7 jobs. Agents run in a browser control center or terminal, not an AI editor; no tab completion. Always-on agents need your own server or OpenHands Cloud.

No: Code in an AI-first editor Not covered
No: Multi-line tab completion and next-edit prediction Not covered
Partial: Let an agent plan and edit across many files Partial replacement
Partial: Run terminal commands and tests from the agent Partial replacement
No: Ask questions about your whole codebase Not covered
Partial: Connect tools and data through MCP Partial replacement
Partial: Run agents in the cloud in the background Partial replacement

For code in an AI-first editor, multi-line tab completion and next-edit prediction, ask questions about your whole codebase, see other Cursor alternatives. You may also need to build an interface, an API server and hosting yourself.

Facts, with sources

Every answer links to where we found it. We read the documentation; we did not install it.

Report a correction
QuestionAnswerSourceChecked
Runs on my machine? Confirmed: Yes; Agent Canvas and the agent server run on your machine by default, or on your own Docker host, VM or Kubernetes cluster It runs locally on your machine by default github.com · OpenHands#readme ↗ today
Code licence Confirmed: MIT Permission is hereby granted, free of charge, to any person obtaining a copy github.com · LICENSE ↗ today
Model weights licence Confirmed: No separate model weights github.com · OpenHands#readme ↗ today
Commercial use Confirmed: Yes, MIT allows commercial use to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell github.com · LICENSE ↗ today
Paid services needed? Partial: Not required with local models (LM Studio, Ollama, vLLM, SGLang); cloud model APIs and OpenHands Cloud are optional Connect OpenHands to local LLM servers such as LM Studio, Ollama, vLLM, and SGLang. docs.openhands.dev · local-llms ↗ today
Hardware Partial: No requirements given for Agent Canvas itself; the recommended local model needs a 24GB GPU or an Apple Silicon Mac with 64GB memory A recent GPU with at least 24GB of VRAM for quantized variants docs.openhands.dev · local-llms ↗ today
Operating systems Confirmed: macOS, Linux and Windows (PowerShell or WSL2); the CLI and per-conversation Docker sandboxes need WSL on Windows For the most reliable local setup, use macOS, Linux, Windows with PowerShell, or Windows with WSL2. docs.openhands.dev · setup ↗ today
Setup effort Confirmed: Medium; install Node.js 24 and uv then the npm package, or run the Docker image, then add a model Install Node.js 24 or later and uv, then verify both tools are available docs.openhands.dev · setup ↗ today

No licence changes since we started watching. Licence watch: follow OpenHands and we email you if its licence terms change.

Growth and activity in detail

Is it popular, and is anyone still working on it? From GitHub.

Numbers appear after the first nightly fetch from GitHub.

Cookie preferences

Choose what we may store in your browser. You can change this any time from Cookie settings in the footer.

Essential

Security, and remembering this choice. rq-consent · 12 months

Always on