OpenHands
Our verdict
The go-to open-source option if you want Devin-style agents on your own infrastructure. More setup than a hosted tool, but you control models, data and sandbox.
- Who it's for
- Teams that need self-hosting or want to customise the agent
- Leverage
- Autonomous dev agents under your control
- Effort
- High
- Date added
What it does
OpenHands is an open-source project for AI-driven development: agents that modify code, run commands and browse, running in sandboxed environments.
How builders use it
- Run it locally with Docker to evaluate.
- Point it at a repo and an issue; review the resulting changes.
- Use it as a base to build your own internal coding agent.
What to expect
Agent capability without sending code to a closed platform.
Related discoveries
- Vibe-Coding & Prompt-to-AppUse case
Run Bolt's open-source core yourself
The bolt.new repo is public: study or run a prompt-to-app agent locally.
Effort: High
- Coding & Shipping AgentsTool
Cursor Cloud Agents
Run Cursor's agent in the cloud so tasks continue while you work on something else.
Effort: Low
- Coding & Shipping AgentsTool
OpenAI Codex CLI
OpenAI's open-source, lightweight coding agent for your terminal.
Effort: Low
- Coding & Shipping AgentsTool
Codex Cloud: delegate tasks to cloud sandboxes
Send coding tasks to Codex in the cloud, review the results, open PRs.
Effort: Medium
Want more like this?
Practical discoveries, without the noise.