ByHeather MacAvelia·Last verified Jun 27, 2026

OpenHands

Name: OpenHands
Author: All Hands AI

★4.6/ 5

by All Hands AI

MCPIndependently Reviewed

Visit openhands.dev →

Open-source autonomous coding agent that writes code, runs commands, and completes GitHub issues end-to-end. Free self-hosted (MIT), free cloud SaaS (BYOK), Enterprise custom. {{github_stars}} GitHub stars.

From

Custom

free ↗

GitHub

⭐ 79.2k

View on GitHub ↗

—

View on G2 ↗

MCP

⚡ Yes

Compatible

OpenHands (formerly OpenDevin) is an open-source platform for autonomous AI software agents built by OpenHands. It enables AI agents to act as full software engineers: writing code, running terminal commands, browsing the web for documentation, interacting with APIs, and submitting pull requests. Given a GitHub issue or task description, OpenHands plans and executes the full solution autonomously. 79.2k GitHub stars and growing. The agent operates in a sandboxed Docker environment with access to terminal, code editor, browser, and file system. It can install dependencies, run test suites, debug failures, and iterate until the task is complete, replicating the workflow of a human developer. OpenHands supports Claude (Anthropic), GPT-4o (OpenAI), Gemini, and open-source models including Llama and Qwen. The platform is consistently benchmarked on SWE-bench with published results, giving teams objective performance data not available from most commercial coding agents. Three deployment options are available. Open Source (free, self-hosted, unlimited conversations, full MIT license, all major LLMs supported via BYOK). Individual SaaS (free, BYOK or at-cost model pricing, GitHub, GitLab, and Bitbucket integration, Jira and Slack integration, unlimited daily conversations). Enterprise (custom pricing, self-hosted VPC option via Kubernetes, Agent Control Plane, unlimited concurrent conversations, Large Codebase SDK, SAML and SSO, multi-user RBAC, budgeting enforcement, usage reporting, priority support with named Customer Engineer). The primary interface is now Agent Canvas, a browser-based UI and backend server that replaced the legacy CLI and Local GUI as the default development surface. Sub-Agent Delegation via TaskToolSet enables multi-agent workflows where different agents handle specialized aspects of complex tasks. MCP support is included across all plans. Integration depth is write-level into the codebase. OpenHands reads the repository, makes changes across multiple files, runs tests, and creates pull requests. Cloud plans add Jira and Slack workflow integrations for issue tracking and notification. The Agent Control Plane on Enterprise enables orchestration of multiple OpenHands agents across different tasks simultaneously. The Cloud Agent SDK provides programmatic access for automation and scripting workflows. Pricing transparency is partial. The free self-hosted and Individual SaaS tiers are clearly defined with unlimited conversations. Enterprise pricing requires a sales conversation and is not published. This is appropriate given the deployment complexity and infrastructure requirements of enterprise-scale autonomous coding agents. OpenHands is not the right fit for teams needing unsupervised production deployment without strong code review processes. It can produce incorrect or partial solutions on ambiguous requirements. Treat it as a highly capable junior engineer that requires review before merging. Devin and GitHub Copilot are commercial alternatives with managed infrastructure and SLAs but at materially higher per-seat costs. Teams not comfortable managing model API costs or Docker infrastructure will find the self-hosted path more demanding than fully managed commercial coding agents.

Pricing

free

View pricing ↗

Segment

b2b

Setup

moderate

Verified

Jun 27, 2026

Transparency

Partial

Contract

Month-to-month

Data training

Opt-out

Autonomy

Human Optional

Capabilities

agentic-codingterminal-agentautonomousopen-sourcebyokweb-search

Pros & Limitations

Editorial assessment

Pros

✓Full autonomous software engineering loop: reads issues, writes code, runs tests, debugs failures, and submits PRs without per-step human direction.
✓Open source with MIT license and BYOK model support: self-host on your own infrastructure with your choice of Claude, GPT-4o, Gemini, or open-source models. Zero vendor lock-in and full code transparency.
✓79.2k GitHub stars and active research community: consistently benchmarked on SWE-bench with published results, giving teams objective performance data not available from most commercial coding agents.

Limitations

⚠Requires careful task scoping and human review: OpenHands works best on bounded, well-specified issues and can produce incorrect or partial solutions on ambiguous requirements. Unsupervised production deployment is risky without strong code review processes.
⚠Enterprise is the only path to advanced team features: multi-user RBAC, centralized billing, SAML/SSO, and the Large Codebase SDK all require custom Enterprise pricing, which is not published.
⚠Self-hosting adds infrastructure overhead: running OpenHands at team scale requires Docker management, compute provisioning, and API key management that adds operational work compared to fully managed commercial alternatives like Devin.

Technical Details

Deployment

cloudself-hosted

Model architectureMulti-model (Claude, GPT-4o, Gemini, Llama, Qwen)

Avg setup timeCloud: under 5 minutes (app.all-hands.dev, connect GitHub). Self-hosted: 30 to 60 minutes (Docker setup, model API key configuration)

Autonomous rateOpenHands autonomously completes software engineering tasks end-to-end: writes code, runs tests, debugs failures, and submits PRs without per-step human approval once the issue or task is assigned. The agent iterates until completion or until it reaches a defined stopping condition.

MCP compatibleYes

Integrations

GitHubGitLabBitbucketJiraSlackLinear