OpenHands

OpenHands is an open source platform for building and running AI coding agents that execute real engineering work instead of only suggesting code. Its homepage promise is to build a workflow once and let agents keep it running: you start with a template or build your own, and trigger agent workflows from GitHub, Slack, Jira, Linear, schedules, and supported integrations, so engineering work keeps moving without someone manually starting every task. The automations its homepage leads with cover the repetitive parts of the software lifecycle: handling engineering requests raised in Slack, triaging new Linear issues and Jira incidents, reviewing every pull request, drafting incident retrospectives, remediating security findings by updating code and opening a pull request, and fixing CI failures by inspecting logs and proposing a fix. Each workflow runs the same sequence, from signal to plan, execution, validation, shipping, and monitoring, while engineers stay in control of review, approval, and merge. OpenHands is multi backend by design: teams can standardize how agent workflows run without standardizing on one agent or model, using Claude Code, Codex, Gemini CLI, OpenHands, or their own stack, with closed models and open weights running locally, in the cloud, or inside a VPC, and MCP connects internal tools, APIs, and data sources. For platform and security teams, its Agent Control Plane gives centralized control over who can run agents, what they can access, where they execute, what they spend, and what they change, and its FAQ stresses that agents run inside your own environment, on premises or in a private cloud, in isolated execution environments with full auditability, so code never leaves your control. It is built for large and legacy codebases as well as new ones, mapping dependencies so multiple agents can work in parallel in the right order. Its homepage reports 93.5K GitHub stars and 597 contributors, with an MIT core licence. Pick OpenHands when your team wants autonomous coding agents wired into its existing engineering stack, on infrastructure and models you choose.