One agent engine, five interfaces, and a license split

What it does OpenHands is a Python-based agentic development toolkit. It exposes the same engine through multiple frontends: a composable library for defining agents in code, a terminal interface, a local React GUI with a REST API, and a hosted cloud version. There is also an enterprise tier for self-hosted Kubernetes deployments with RBAC, multi-user support, and Slack, Jira, and Linear integrations.

The interesting bit The project draws a hard line between its MIT-licensed core and a source-available enterprise directory that requires a paid license after one month—a rare dual-life arrangement in the open-source agent world. It also publicly claims a 77.6 SWE-bench score, suggesting the maintainers care about measurable benchmark performance rather than vibe-based demos.

Key highlights

• Composable Python SDK that scales from local scripts to thousands of cloud agents
• Terminal experience comparable to Claude Code or Codex, pluggable with Claude, GPT, or other LLMs
• Local GUI bundles a React SPA and REST API for laptop-based agent workflows
• Cloud and Enterprise tiers add multi-user support, RBAC, Jira/Linear/Slack integrations, and VPC self-hosting via Kubernetes
• Core Docker images and SDK fully MIT-licensed; enterprise code is visible but gated

Caveats

• Enterprise features are source-available but require a paid license after one month of self-hosted use
• The README showcases major corporate logos without explaining what “trusted by” actually means in practice

Verdict Developers evaluating AI coding agents should look here if they want an open-core toolkit with multiple interfaces and a credible benchmark score; teams allergic to license complexity or looking for a simple, single-purpose tool should probably keep scrolling.