OpenHands
Agent Canvas: a self-hosted hub to run OpenHands, Claude Code, Codex and other coding agents with automations.
Run OpenHands or any ACP agent locally, in Docker or on a VM, and automate work via Slack and GitHub.
The OpenHands repository now holds Agent Canvas, a web interface for starting conversations with coding agents and scheduling automations. It suits developers and small engineering teams who want agents running on their own machine or server with a model key of their choice.
What it is
Agent Canvas runs the open source OpenHands agent by default and can also drive Claude Code, Codex, Gemini CLI, Pi, OpenCode or any agent speaking the Agent Client Protocol. It connects to local, Docker, VM or OpenHands Cloud backends, and automations can run on a schedule or on webhook events from Slack, GitHub and Linear.
Who it's for
It suits developers comfortable with Node.js, uv and Docker who want agents working on their repositories. It does not suit anyone expecting a one-click app: the no-sandbox mode gives the agent full access to your filesystem.
Key features
Multiple agent backends
Switch between local, Docker, VM and cloud agent servers from the same interface.
Any ACP agent
Run OpenHands, Claude Code, Codex, Gemini, Pi, OpenCode or other Agent Client Protocol agents.
Scheduled automations
Run agents on a schedule or on webhooks, such as turning GitHub issues into tasks or posting reports to Slack.
Docker sandbox option
Each conversation can run in its own Docker container with its workspace mounted.
Bring your own model
Use any LLM, with saved LLM profiles.
Separate agent server
The agent runtime is a REST Agent Server you can host on a shared machine for a team.
Up and running in 3 steps
Our honest verdict
What's great
- You can pick agents and models freely instead of being tied to one vendor's coding assistant.
- Automations tie agents to Slack, GitHub, Linear and other services without writing glue code.
- A Docker sandbox mode isolates conversation execution from your host machine.
- Frontend, SDK, client and automation service live in separate repositories with documented boundaries, which helps if you want to extend it.
Where it falls short
- The quickstart without a sandbox lets the agent access your whole filesystem, and the README warns about this.
- You need Node.js 24 or later and uv, and Docker for sandboxing; a server deployment needs security hardening per SELF_HOSTING.md.
- The free tier is single user or shared identity; multi-user identity, RBAC and SSO require the paid Teams plan.
- Your own LLM key is required, and agent runs spend model tokens you pay for.
Pricing & license
Agent Canvas is free under the MIT licence for self-hosting with a single or shared identity, plus your own LLM costs. OpenHands' pricing page lists Teams at $500/month (10 users included, $50/month per extra seat, SSO and RBAC, 5,000 sessions a month), and Enterprise on custom pricing; OpenHands Cloud shows no price there (https://openhands.dev/pricing).
Open remote AI-training roles
860 open remote AI-training roles, updated November 14, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Is OpenHands still the coding agent I remember?
The repository README now presents Agent Canvas, a control center that runs the open source OpenHands agent by default and can run other agents. The agent server and SDK live in the software-agent-sdk repository.
Is it safe to run on my laptop?
The README warns that running without a sandbox gives the agent full access to your filesystem. Use the Docker sandbox options, and see SELF_HOSTING.md for server hardening.
What do I need besides the software?
A model API key, since you bring your own LLM. Teams features such as SSO need a paid plan.
