Zero Setup AI Coding with OpenHands

Agent Canvas is the browser client and control center for OpenHands conversations and automations. You start sessions, watch the agent work step by step, and branch off a previous message when you want to try a different path. It connects to a local, self-hosted, cloud or enterprise backend, so the same interface follows you across setups. For anyone who wants to skip local installation, this is the fastest way in.

OpenHands Cloud is the managed service that removes the sandbox work. Hosted execution, integrations, access controls and budget management all run on OpenHands infrastructure, so you connect a repository and go. Usage reporting and spending limits keep an eye on cost as a team grows. It is the option for people who want the agent working today rather than after a Docker install.

The OpenHands SDK is a composable Python library for building agents that work with code. It ships the reasoning loop, tools, conversations and workspaces as reusable parts, and the same repo includes Agent Server, which exposes agent execution through REST and WebSocket APIs. Teams building their own products use it to embed agentic development instead of inventing the loop. Custom tools and file-based sub-agents extend it without a rewrite.

Skills give the agent specialized behavior through structured prompts, triggered by keywords, file paths or the repository itself. Instead of repeating the same instructions every session, you write a skill once for how your team reviews code or upgrades a library. Global skills shared across users and organization-level skills keep policy consistent. It is the layer that turns a general agent into one that follows your house rules.

Automations run agent work on a schedule or on an event, and OpenHands wires into GitHub, Slack and Jira so the right agent fires when something happens. An issue opens, an agent picks it up, does the work and raises a pull request, all logged and governed by the same cost and security policies as a manual session. Automated QA testing runs the software rather than only reading the diff. That event-to-PR path is where repetitive chores stop landing on a person.

OpenHands Enterprise adds the governance layer for teams running agents at scale. It self-hosts in your VPC via Kubernetes with RBAC, an LLM gateway, budgeting and observability, all source-available. You keep control over which models run, which data they touch and who can start a task. For regulated teams, that control is the difference between a pilot and a rollout.

OpenHands connects to any model supported by LiteLLM, covering OpenAI, Anthropic, Google, AWS Bedrock, Azure, Groq and OpenRouter among others. You can point it at a local server such as Ollama, vLLM or LM Studio, or use the OpenHands provider for agentic coding models. LLM profiles let you switch models mid-conversation or route cheap work to a smaller one. That range matters when data rules or budget push you off a single vendor.