# Best AI coding agents for Open Source (2026)

> The best AI coding agents platforms for open source: DeepSeek Harness (dsh) leads, for teams that want to read, fork or self-host the agent itself: the.

For open source, **DeepSeek Harness (dsh)** is our pick: The license decides this. Teams that want to read, fork or self-host the agent itself: the license, which model providers it can call, and what a hosted option costs if you would rather not run it yourself. Below is the full ranking and the tradeoffs, or read [how we score](https://www.versusref.com/methodology/).

## What matters for open source

Weight ×5 = decisive, ×1 = relevant.

| Fact | Weight | DeepSeek Harness (dsh) | Freebuff | Gemini CLI | Goose | Grok Build |
| --- | --- | --- | --- | --- | --- | --- |
| License | ×5 | MIT (Oct 2) | Apache-2.0 (Oct 2) | Apache-2.0 (Oct 2) | Apache-2.0 (Oct 2) | Apache-2.0 (Oct 2) |
| Supported model providers | ×4 | n/a | Curated hosted catalog (Oct 2) | Google only: Gemini Code Assist (Oct 2) | 45+ providers incl. Anthropic, OpenAI, Gemini +20 more (Oct 2) | xAI (Grok models); custom models via base URL + API key (Oct 2) |
| Runs local models | ×3 | n/a | n/a | n/a | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) |
| GitHub stars | ×2 | 242,700 (Oct 3) | 13,100 (Oct 2) | 107,200 (Oct 2) | ~54,870 (Oct 2) | 27,200 (Oct 2) |
| Hosted / paid option | ×2 | n/a | Codebuff paid subscriptions: $100/mo (1x usage) (Oct 2) | Gemini Code Assist Standard $19-22.80/user/mo (Oct 2) | false (Oct 2) | n/a |

- ×5 **License:** MIT and Apache-2.0 allow almost anything; source-available licenses like FSL restrict competing commercial use. The license decides what you may actually ship.
- ×4 **Supported model providers:** An open agent is only as useful as the models it can drive; a wide provider list keeps you off any one vendor's price list.
- ×3 **Runs local models:** Local-model support is what makes an open-source agent fully self-contained.
- ×2 **GitHub stars:** Stars are a rough proxy for community size, which decides how fast bugs get fixed and integrations appear.
- ×2 **Hosted / paid option:** A paid hosted tier is the escape hatch when running the agent yourself stops being worth the time.

## The ranking, tool by tool

| Rank | Tool | Verdict | Score | Price |
| --- | --- | --- | --- | --- |
| 1 | [DeepSeek Harness (dsh)](https://www.versusref.com/coding-agents/tools/deepseek-harness/) (OSS) | The license decides this. | 2 of 2 points · 1 matchup | See pricing |
| 2 | [Freebuff](https://www.versusref.com/coding-agents/tools/freebuff/) (OSS) | For teams that want to read or fork the agent, the license matters most. | 2 of 2 points · 1 matchup | See pricing |
| 3 | [Gemini CLI](https://www.versusref.com/coding-agents/tools/gemini-cli/) (OSS) | For teams that want to read, fork or self-host the agent, the license matters most, and Gemini CLI is Apache-2.0 while Google Antigravity does not publish a license. | 2 of 2 points · 1 matchup | See pricing |
| 4 | [Goose](https://www.versusref.com/coding-agents/tools/goose/) (OSS) | For teams that want to read, fork or self-host the agent, the license comes first. | 2 of 2 points · 1 matchup | See pricing |
| 5 | [Grok Build](https://www.versusref.com/coding-agents/tools/grok-build/) (OSS) | License decides this, and Grok Build is Apache-2.0 while Claude Code is proprietary, with its npm package pointing to a license in its README. | 2 of 2 points · 1 matchup | See pricing |
| 6 | [Kimi Code](https://www.versusref.com/coding-agents/tools/kimi-code/) (OSS) | License decides this one. | 2 of 2 points · 1 matchup | See pricing |
| 7 | [Letta Code](https://www.versusref.com/coding-agents/tools/letta-code/) (OSS) | The license decides this. | 2 of 2 points · 1 matchup | See pricing |
| 8 | [Mistral Vibe](https://www.versusref.com/coding-agents/tools/mistral-vibe/) (OSS) | The license is the deciding factor for teams that want to read or fork the agent. | 2 of 2 points · 1 matchup | See pricing |
| 9 | [OpenHands](https://www.versusref.com/coding-agents/tools/openhands/) (OSS) | The license decides this, and only OpenHands has an open one: it is MIT licensed with 89,799 GitHub stars. | 2 of 2 points · 1 matchup | See pricing |
| 10 | [Pi](https://www.versusref.com/coding-agents/tools/pi/) (OSS) | The license is the deciding factor: Pi is MIT licensed, while Claude Code is proprietary, pointing to a license in its README on the npm package. | 2 of 2 points · 1 matchup | See pricing |
| 11 | [Qwen Code](https://www.versusref.com/coding-agents/tools/qwen-code/) (OSS) | The license matters most here. | 2 of 2 points · 1 matchup | See pricing |
| 12 | [Tabby](https://www.versusref.com/coding-agents/tools/tabby/) (OSS) | The license is what matters most here. | 2 of 2 points · 1 matchup | See pricing |
| 13 | [Warp](https://www.versusref.com/coding-agents/tools/warp/) | The license is what matters most here, and it is decisive. | 2 of 2 points · 1 matchup | See pricing |
| 14 | [Zed](https://www.versusref.com/coding-agents/tools/zed/) (OSS) | The license matters most here, and Zed is licensed GPL-3.0-or-later with Apache-2.0 components, while Cursor publishes no license. | 2 of 2 points · 1 matchup | See pricing |
| 15 | [OpenCode](https://www.versusref.com/coding-agents/tools/opencode/) (OSS) | The license matters most here. | 5 of 6 points · 3 matchups | See pricing |
| 16 | [OpenAI Codex](https://www.versusref.com/coding-agents/tools/openai-codex/) | For teams that want to read or fork the agent, the license matters most. | 6 of 8 points · 4 matchups | See pricing |
| 17 | [Aider](https://www.versusref.com/coding-agents/tools/aider/) (OSS) | The license decides this: Aider is Apache-2.0 while Claude Code is proprietary, so only Aider can be read, forked and self-hosted as source. | 4 of 6 points · 3 matchups | See pricing |
| 18 | [Cline](https://www.versusref.com/coding-agents/tools/cline/) (OSS) | The license matters most here. | 2 of 4 points · 2 matchups | See pricing |
| 19 | [Command Code](https://www.versusref.com/coding-agents/tools/command-code/) | Neither is open source: Claude Code is proprietary, and Command Code's npm package declares itself unlicensed. | 1 of 2 points · 1 matchup | See pricing |
| 20 | [Factory](https://www.versusref.com/coding-agents/tools/factory/) | Neither Devin nor Factory publishes a license, a list of supported model providers, GitHub stars or a separate hosted option, so the main open source signals are missing on both sides. | 1 of 2 points · 1 matchup | See pricing |
| 21 | [Google Antigravity](https://www.versusref.com/coding-agents/tools/google-antigravity/) | Neither tool is a clear open-source choice. | 2 of 6 points · 3 matchups | See pricing |
| 22 | [GitHub Copilot](https://www.versusref.com/coding-agents/tools/github-copilot/) | For teams that want to read, fork or self-host the agent, the license matters most, and neither Cursor nor GitHub Copilot publishes one. | 4.5 of 16 points · 8 matchups | See pricing |
| 23 | [Kilo Code](https://www.versusref.com/coding-agents/tools/kilo-code/) (OSS) | Both licenses let you read, fork and self-host: Cline is Apache-2.0 and Kilo Code is MIT, so the license does not separate them. | 1 of 4 points · 2 matchups | See pricing |
| 24 | [Kiro](https://www.versusref.com/coding-agents/tools/kiro/) | Spec-driven AI IDE + CLI (ex-Q Developer CLI) + web/cloud agent; AWS successor to Amazon Q Developer. | 1 of 4 points · 2 matchups | See pricing |
| 25 | [Amp](https://www.versusref.com/coding-agents/tools/amp/) | Frontier multi-model agent + cloud dev environments (orbs); also terminal_agent (Amp CLI, connects to VS Code/Cursor/Windsurf/Zed/Neovim). OWNERSHIP: no longer branded as Sourcegraph - operated by "Amp Frontier Corporation" (privacy policy, modified Sep 10, 2026); about page: "We are an independent agent research lab"; CLI npm package moved from @sourcegraph/amp to @ampcode/cli (May 14, 2026 changelog). Free Hobby tier with BYOK/ChatGPT-sub since Sep 13, 2026. | 0.5 of 2 points · 1 matchup | See pricing |
| 26 | [Augment Code](https://www.versusref.com/coding-agents/tools/augment-code/) | Context-engine coding platform for large codebases; now led by Cosmos cloud agents + Auggie CLI (also ide_extension: VS Code, JetBrains, Vim/Neovim chat). Brief seeded it as an IDE extension; the 2026 pricing page leads with "Try Cosmos". Flat team pricing (up to 50 seats, no per-seat charge) with a pooled $ usage balance. | 0.5 of 2 points · 1 matchup | See pricing |
| 27 | [Cosine](https://www.versusref.com/coding-agents/tools/cosine/) | "The Sovereign AI Lab": coding agent + own post-trained Lumen models, pitched at regulated/air-gapped orgs and niche languages (COBOL, Fortran, Verilog). Formerly branded Genie (Genie CLI introduced Q2 2025). Secondary segment: terminal_agent (cos CLI). | 0.5 of 2 points · 1 matchup | See pricing |
| 28 | [GitLab Duo Agent Platform](https://www.versusref.com/coding-agents/tools/gitlab-duo-agent-platform/) | DevSecOps-native agents and flows for GitLab Premium/Ultimate; also IDE extensions (VS Code, JetBrains) and a Duo CLI; self-managed and air-gapped self-hosted-model options. | 0.5 of 2 points · 1 matchup | See pricing |
| 29 | [JetBrains Junie](https://www.versusref.com/coding-agents/tools/junie/) | JetBrains IDE agent, now also a terminal agent (Junie CLI) and runs in JetBrains Air; metered in JetBrains AI Credits ($1 each) with BYOK at provider rates, zero markup. | 0.5 of 2 points · 1 matchup | See pricing |
| 30 | [Tabnine](https://www.versusref.com/coding-agents/tools/tabnine/) | Private, governed enterprise coding agent with self-hosted and air-gapped deployment. Now a Tricentis product (acquired 2026-07-30). Secondary segment: terminal_agent (Tabnine Plugin for OpenCode). | 0.5 of 2 points · 1 matchup | See pricing |
| 31 | [Trae](https://www.versusref.com/coding-agents/tools/trae/) | Low-priced AI IDE (Free/Lite/Pro/Pro+/Ultra); product family now split into TraeCode (IDE) and TraeWork (web/desktop/mobile workspace built on SOLO). Owner per brief: ByteDance - the legal entity could not be read from a primary source (privacy policy/ToS are client-rendered); enterprise page cites Douyin as a customer. Chats may be used for training unless Privacy Mode is on. | 0.5 of 2 points · 1 matchup | See pricing |
| 32 | [Verdent](https://www.versusref.com/coding-agents/tools/verdent/) | Plan-Code-Verify multi-model agent; desktop app (primary, most changelog activity), VS Code/JetBrains plugins, Slack/Telegram control; increasingly pitched as an 'AI technical cofounder' with app deployment. | 0.5 of 2 points · 1 matchup | See pricing |
| 33 | [Devin Desktop (formerly Windsurf)](https://www.versusref.com/coding-agents/tools/devin-desktop/) | Agentic AI IDE (ex-Windsurf) inside the Devin family; shares plans with Devin Cloud and Devin CLI. | 1 of 6 points · 3 matchups | See pricing |
| 34 | [Cursor](https://www.versusref.com/coding-agents/tools/cursor/) | AI IDE with CLI and cloud agents; ships its own Grok/Composer models. Acquired by SpaceX (completed 2026-08-14, per cursor.com/blog/joining-spacex); also segments terminal_agent, cloud_agent. | 3 of 22 points · 11 matchups | See pricing |
| 35 | [Devin](https://www.versusref.com/coding-agents/tools/devin/) | Autonomous cloud coding agent + terminal agent (Devin CLI); plans shared with Devin Desktop. | 1 of 8 points · 4 matchups | See pricing |
| 36 | [Claude Code](https://www.versusref.com/coding-agents/tools/claude-code/) | Terminal-first coding agent included in Claude Pro/Max/Team/Enterprise plans or billed via API; also IDE extension, desktop app and cloud sessions (cloud_agent). | 1.5 of 38 points · 19 matchups | See pricing |
| 37 | [Crush](https://www.versusref.com/coding-agents/tools/crush/) (OSS) | Source-available (FSL, converts to MIT after 2 years) BYO-model terminal agent; optional Charm-hosted 'Hyper' inference subscription. | 0 of 2 points · 1 matchup | See pricing |
| 38 | [gptme](https://www.versusref.com/coding-agents/tools/gptme/) (OSS) | MIT local-first terminal agent; also web UI, desktop app, REST server, MCP/ACP; persistent autonomous-agent template. | 0 of 2 points · 1 matchup | See pricing |

### 1. DeepSeek Harness (dsh)

The license decides this. [Full DeepSeek Harness (dsh) vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-deepseek-harness/)

### 2. Freebuff

For teams that want to read or fork the agent, the license matters most. [Full Freebuff vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-freebuff/)

### 3. Gemini CLI

For teams that want to read, fork or self-host the agent, the license matters most, and Gemini CLI is Apache-2.0 while Google Antigravity does not publish a license. [Full Gemini CLI vs Google Antigravity verdict](https://www.versusref.com/coding-agents/gemini-cli-vs-google-antigravity/)

### 4. Goose

For teams that want to read, fork or self-host the agent, the license comes first. [Full Goose vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-goose/)

### 5. Grok Build

License decides this, and Grok Build is Apache-2.0 while Claude Code is proprietary, with its npm package pointing to a license in its README. [Full Grok Build vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-grok-build/)

### 6. Kimi Code

License decides this one. [Full Kimi Code vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-kimi-code/)

### 7. Letta Code

The license decides this. [Full Letta Code vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-letta-code/)

### 8. Mistral Vibe

The license is the deciding factor for teams that want to read or fork the agent. [Full Mistral Vibe vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-mistral-vibe/)

### 9. OpenHands

The license decides this, and only OpenHands has an open one: it is MIT licensed with 89,799 GitHub stars. [Full OpenHands vs Devin verdict](https://www.versusref.com/coding-agents/devin-vs-openhands/)

### 10. Pi

The license is the deciding factor: Pi is MIT licensed, while Claude Code is proprietary, pointing to a license in its README on the npm package. [Full Pi vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-pi/)

### 11. Qwen Code

The license matters most here. [Full Qwen Code vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-qwen-code/)

### 12. Tabby

The license is what matters most here. [Full Tabby vs GitHub Copilot verdict](https://www.versusref.com/coding-agents/github-copilot-vs-tabby/)

### 13. Warp

The license is what matters most here, and it is decisive. [Full Warp vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-warp/)

### 14. Zed

The license matters most here, and Zed is licensed GPL-3.0-or-later with Apache-2.0 components, while Cursor publishes no license. [Full Zed vs Cursor verdict](https://www.versusref.com/coding-agents/cursor-vs-zed/)

### 15. OpenCode

The license matters most here. [Full OpenCode vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-opencode/)

Both are MIT licensed, so the license does not separate them. [Full OpenCode vs Kilo Code verdict](https://www.versusref.com/coding-agents/kilo-code-vs-opencode/)

For teams that want to read, fork or self-host the agent, the license matters most. [Full OpenCode vs Crush verdict](https://www.versusref.com/coding-agents/crush-vs-opencode/)

### 16. OpenAI Codex

For teams that want to read or fork the agent, the license matters most. [Full OpenAI Codex vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-openai-codex/)

For teams that want to read or fork the agent, the license is what matters most. [Full OpenAI Codex vs Cursor verdict](https://www.versusref.com/coding-agents/cursor-vs-openai-codex/)

The license is what matters most for teams that want to read or fork the agent. [Full OpenAI Codex vs GitHub Copilot verdict](https://www.versusref.com/coding-agents/github-copilot-vs-openai-codex/)

### 17. Aider

The license decides this: Aider is Apache-2.0 while Claude Code is proprietary, so only Aider can be read, forked and self-hosted as source. [Full Aider vs Claude Code verdict](https://www.versusref.com/coding-agents/aider-vs-claude-code/)

Both carry the Apache-2.0 license, but for OpenAI Codex it covers the Codex CLI, while Aider's whole tool is free and open source. [Full Aider vs OpenAI Codex verdict](https://www.versusref.com/coding-agents/aider-vs-openai-codex/)

Both are permissively licensed, Aider under Apache-2.0 and gptme under MIT, and both run local models. [Full Aider vs gptme verdict](https://www.versusref.com/coding-agents/aider-vs-gptme/)

### 18. Cline

The license matters most here. [Full Cline vs Cursor verdict](https://www.versusref.com/coding-agents/cline-vs-cursor/)

### 19. Command Code

Neither is open source: Claude Code is proprietary, and Command Code's npm package declares itself unlicensed. [Full Command Code vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-command-code/)

### 20. Factory

Neither Devin nor Factory publishes a license, a list of supported model providers, GitHub stars or a separate hosted option, so the main open source signals are missing on both sides. [Full Factory vs Devin verdict](https://www.versusref.com/coding-agents/devin-vs-factory/)

### 21. Google Antigravity

Neither tool is a clear open-source choice. [Full Google Antigravity vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-google-antigravity/)

Neither tool publishes a license, a list of supported model providers, GitHub stars or a separate hosted option, so the main open source signals are missing on both sides. [Full Google Antigravity vs Cursor verdict](https://www.versusref.com/coding-agents/cursor-vs-google-antigravity/)

### 22. GitHub Copilot

For teams that want to read, fork or self-host the agent, the license matters most, and neither Cursor nor GitHub Copilot publishes one. [Full GitHub Copilot vs Cursor verdict](https://www.versusref.com/coding-agents/cursor-vs-github-copilot/)

Neither tool is an open-source agent you can fork. [Full GitHub Copilot vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-github-copilot/)

The license matters most to teams that want to read or fork the agent, and neither Devin Desktop nor GitHub Copilot publishes one, nor GitHub stars or a supported provider list. [Full GitHub Copilot vs Devin Desktop (formerly Windsurf) verdict](https://www.versusref.com/coding-agents/devin-desktop-vs-github-copilot/)

### 23. Kilo Code

Both licenses let you read, fork and self-host: Cline is Apache-2.0 and Kilo Code is MIT, so the license does not separate them. [Full Kilo Code vs Cline verdict](https://www.versusref.com/coding-agents/cline-vs-kilo-code/)

### 24. Kiro

Spec-driven AI IDE + CLI (ex-Q Developer CLI) + web/cloud agent; AWS successor to Amazon Q Developer. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 25. Amp

Frontier multi-model agent + cloud dev environments (orbs); also terminal_agent (Amp CLI, connects to VS Code/Cursor/Windsurf/Zed/Neovim). OWNERSHIP: no longer branded as Sourcegraph - operated by "Amp Frontier Corporation" (privacy policy, modified Sep 10, 2026); about page: "We are an independent agent research lab"; CLI npm package moved from @sourcegraph/amp to @ampcode/cli (May 14, 2026 changelog). Free Hobby tier with BYOK/ChatGPT-sub since Sep 13, 2026. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 26. Augment Code

Context-engine coding platform for large codebases; now led by Cosmos cloud agents + Auggie CLI (also ide_extension: VS Code, JetBrains, Vim/Neovim chat). Brief seeded it as an IDE extension; the 2026 pricing page leads with "Try Cosmos". Flat team pricing (up to 50 seats, no per-seat charge) with a pooled $ usage balance. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 27. Cosine

"The Sovereign AI Lab": coding agent + own post-trained Lumen models, pitched at regulated/air-gapped orgs and niche languages (COBOL, Fortran, Verilog). Formerly branded Genie (Genie CLI introduced Q2 2025). Secondary segment: terminal_agent (cos CLI). No won verdicts for this use case yet; it ranks on ties and near-misses.

### 28. GitLab Duo Agent Platform

DevSecOps-native agents and flows for GitLab Premium/Ultimate; also IDE extensions (VS Code, JetBrains) and a Duo CLI; self-managed and air-gapped self-hosted-model options. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 29. JetBrains Junie

JetBrains IDE agent, now also a terminal agent (Junie CLI) and runs in JetBrains Air; metered in JetBrains AI Credits ($1 each) with BYOK at provider rates, zero markup. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 30. Tabnine

Private, governed enterprise coding agent with self-hosted and air-gapped deployment. Now a Tricentis product (acquired 2026-07-30). Secondary segment: terminal_agent (Tabnine Plugin for OpenCode). No won verdicts for this use case yet; it ranks on ties and near-misses.

### 31. Trae

Low-priced AI IDE (Free/Lite/Pro/Pro+/Ultra); product family now split into TraeCode (IDE) and TraeWork (web/desktop/mobile workspace built on SOLO). Owner per brief: ByteDance - the legal entity could not be read from a primary source (privacy policy/ToS are client-rendered); enterprise page cites Douyin as a customer. Chats may be used for training unless Privacy Mode is on. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 32. Verdent

Plan-Code-Verify multi-model agent; desktop app (primary, most changelog activity), VS Code/JetBrains plugins, Slack/Telegram control; increasingly pitched as an 'AI technical cofounder' with app deployment. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 33. Devin Desktop (formerly Windsurf)

Agentic AI IDE (ex-Windsurf) inside the Devin family; shares plans with Devin Cloud and Devin CLI. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 34. Cursor

AI IDE with CLI and cloud agents; ships its own Grok/Composer models. Acquired by SpaceX (completed 2026-08-14, per cursor.com/blog/joining-spacex); also segments terminal_agent, cloud_agent. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 35. Devin

Autonomous cloud coding agent + terminal agent (Devin CLI); plans shared with Devin Desktop. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 36. Claude Code

Terminal-first coding agent included in Claude Pro/Max/Team/Enterprise plans or billed via API; also IDE extension, desktop app and cloud sessions (cloud_agent). No won verdicts for this use case yet; it ranks on ties and near-misses.

### 37. Crush

Source-available (FSL, converts to MIT after 2 years) BYO-model terminal agent; optional Charm-hosted 'Hyper' inference subscription. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 38. gptme

MIT local-first terminal agent; also web UI, desktop app, REST server, MCP/ACP; persistent autonomous-agent template. No won verdicts for this use case yet; it ranks on ties and near-misses.

Source: https://www.versusref.com/coding-agents/best/open-source/