# Best AI coding agents for Enterprise (2026)

> The best AI coding agents platforms for enterprise: GitLab Duo Agent Platform leads, for engineering orgs buying for many developers: no training on company.

For enterprise, **GitLab Duo Agent Platform** is our pick: Training policy and self-hosting matter most to a security review. Engineering orgs buying for many developers: no training on company code, SSO and audit logs, a self-hosted or VPC option, and a compliance record a security review will accept. Below is the full ranking and the tradeoffs, or read [how we score](https://www.versusref.com/methodology/).

## What matters for enterprise

Weight ×5 = decisive, ×1 = relevant.

| Fact | Weight | GitLab Duo Agent Platform | Tabnine | Cursor | Claude Code | Devin Desktop (formerly Windsurf) |
| --- | --- | --- | --- | --- | --- | --- |
| Trains on your code? | ×5 | 'GitLab does not train generative AI models.' 'All... (Oct 2) | "Tabnine has a no-train-no-retain policy. This is in... (Oct 2) | Privacy Mode on: not used for training by Cursor or... (Oct 2) | Consumer (Free, Pro, Max): user choice (Oct 2) | Self-serve: "By default (Oct 2) |
| Self-hosted / VPC deployment | ×4 | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) |
| Zero data retention option | ×3 | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) |
| SSO | ×3 | SAML SSO on Premium and Ultimate (GitLab.com groups) (Oct 2) | SAML or OAuth 2.0 / OpenID Connect (one per organization) (Oct 2) | SAML/OIDC SSO on Teams ($40/user/mo) and above (Oct 2) | SSO on Team and Enterprise (Oct 2) | Enterprise only (SAML/OIDC); SCIM (Oct 2) |
| SOC 2 | ×2 | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) |
| Admin audit logs | ×2 | n/a | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) | ✓  Yes (Oct 2) |

- ×5 **Trains on your code?:** Whether the vendor trains on your code (and whether that is opt-in or opt-out) is the first question every security review asks.
- ×4 **Self-hosted / VPC deployment:** A self-hosted or VPC deployment keeps source code inside your own network boundary.
- ×3 **Zero data retention option:** Zero data retention with model providers limits what leaves your control on each request.
- ×3 **SSO:** SSO, and which plan it sits on, decides whether access can be managed centrally.
- ×2 **SOC 2:** A SOC 2 attestation shortens procurement; "aligned with" is not the same thing.
- ×2 **Admin audit logs:** Audit logs let admins see who ran what, which regulated teams need.

## The ranking, tool by tool

| Rank | Tool | Verdict | Score | Price |
| --- | --- | --- | --- | --- |
| 1 | [GitLab Duo Agent Platform](https://www.versusref.com/coding-agents/tools/gitlab-duo-agent-platform/) | Training policy and self-hosting matter most to a security review. | 2 of 2 points · 1 matchup | See pricing |
| 2 | [Tabnine](https://www.versusref.com/coding-agents/tools/tabnine/) | Training on company code is the deciding factor for enterprise buyers. | 2 of 2 points · 1 matchup | See pricing |
| 3 | [Cursor](https://www.versusref.com/coding-agents/tools/cursor/) | Code training policy matters most here. | 19 of 22 points · 11 matchups | See pricing |
| 4 | [Claude Code](https://www.versusref.com/coding-agents/tools/claude-code/) | Claude Code is the only one here with an enterprise record. | 29 of 38 points · 19 matchups | See pricing |
| 5 | [Devin Desktop (formerly Windsurf)](https://www.versusref.com/coding-agents/tools/devin-desktop/) | Neither vendor's training policy, as published, settles the question on its own, so the deployment and control features decide it. | 4 of 6 points · 3 matchups | See pricing |
| 6 | [OpenAI Codex](https://www.versusref.com/coding-agents/tools/openai-codex/) | Neither vendor publishes a crisp training statement: GitHub Copilot's policy note reads "From 2026-04-24" and OpenAI Codex's reads "Business". | 4 of 8 points · 4 matchups | See pricing |
| 7 | [OpenCode](https://www.versusref.com/coding-agents/tools/opencode/) (OSS) | Training on company code and self-hosting matter most here. | 3 of 6 points · 3 matchups | See pricing |
| 8 | [Cline](https://www.versusref.com/coding-agents/tools/cline/) (OSS) | A security review starts with code training and deployment control, and Cline answers both. | 2 of 4 points · 2 matchups | See pricing |
| 9 | [Amp](https://www.versusref.com/coding-agents/tools/amp/) | What matters most here is whether the vendor trains on company code. | 1 of 2 points · 1 matchup | See pricing |
| 10 | [Factory](https://www.versusref.com/coding-agents/tools/factory/) | For a security review, training on company code matters most. | 1 of 2 points · 1 matchup | See pricing |
| 11 | [Gemini CLI](https://www.versusref.com/coding-agents/tools/gemini-cli/) (OSS) | For a security review, the training policy matters most. | 1 of 2 points · 1 matchup | See pricing |
| 12 | [JetBrains Junie](https://www.versusref.com/coding-agents/tools/junie/) | Training policy matters most to a security review, and JetBrains Junie gives the clearer answer: no training on inputs or outputs unless expressly agreed. | 1 of 2 points · 1 matchup | See pricing |
| 13 | [Tabby](https://www.versusref.com/coding-agents/tools/tabby/) (OSS) | Neither tool publishes a clear training commitment: GitHub Copilot's policy note reads "From 2026-04-24" and Tabby publishes none. | 1 of 2 points · 1 matchup | See pricing |
| 14 | [Aider](https://www.versusref.com/coding-agents/tools/aider/) (OSS) | Aider publishes the more reassuring enterprise posture. | 2 of 6 points · 3 matchups | See pricing |
| 15 | [Devin](https://www.versusref.com/coding-agents/tools/devin/) | Training on company code is the first question, and neither gives a clean answer: Devin's self-serve plans may train by default, and OpenHands publishes no training policy. | 2 of 8 points · 4 matchups | See pricing |
| 16 | [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 of 2 points · 1 matchup | See pricing |
| 17 | [Command Code](https://www.versusref.com/coding-agents/tools/command-code/) | Coding agent for open models (DeepSeek, Kimi, GLM, Qwen, MiniMax) with taste learning; CLI + desktop + ACP for Zed; cheap $1-$20 plans with published $ rolling windows. | 0 of 2 points · 1 matchup | See pricing |
| 18 | [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 of 2 points · 1 matchup | See pricing |
| 19 | [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 |
| 20 | [DeepSeek Harness (dsh)](https://www.versusref.com/coding-agents/tools/deepseek-harness/) (OSS) | 'Everything is a Plugin' (Cordis) agent harness from DeepSeek; developer preview with breaking changes; general-purpose, not coding-only. | 0 of 2 points · 1 matchup | See pricing |
| 21 | [Freebuff](https://www.versusref.com/coding-agents/tools/freebuff/) (OSS) | Ad-funded free coding agent on curated low-cost models. History: launched as Codebuff (YC); repo and operator (Freebuff, Inc., legal docs effective 2026-09-02) now present Freebuff, built on the Codebuff open multi-agent framework; codebuff.com still sells the premium-model Codebuff CLI ($100-$500/mo). Also desktop, web builder, cloud IDE. | 0 of 2 points · 1 matchup | See pricing |
| 22 | [GitHub Copilot](https://www.versusref.com/coding-agents/tools/github-copilot/) | Multi-model IDE extension plus Copilot cloud agent (formerly 'coding agent'; cloud_agent) and Copilot CLI (terminal_agent). Billing moved from premium requests to GitHub AI Credits on 2026-06-01. | 0 of 16 points · 8 matchups | See pricing |
| 23 | [Google Antigravity](https://www.versusref.com/coding-agents/tools/google-antigravity/) | Agent-first IDE + CLI + desktop agent manager; Google's consumer coding product since Gemini CLI's 2026-06-18 consumer cutover. | 0 of 6 points · 3 matchups | See pricing |
| 24 | [Goose](https://www.versusref.com/coding-agents/tools/goose/) (OSS) | Local, BYO-model general-purpose agent; also a native desktop app (Rust). Donated by Block to the Linux Foundation's Agentic AI Foundation in 2026. | 0 of 2 points · 1 matchup | See pricing |
| 25 | [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 |
| 26 | [Grok Build](https://www.versusref.com/coding-agents/tools/grok-build/) (OSS) | Subscription-bundled vendor terminal agent (SuperGrok / X Premium+), open-sourced Apache-2.0; also embeds in editors via ACP. | 0 of 2 points · 1 matchup | See pricing |
| 27 | [Kilo Code](https://www.versusref.com/coding-agents/tools/kilo-code/) (OSS) | OSS multi-surface agent; zero-markup inference gateway + Kilo Pass credit bundles; started as a Roo Code fork, now built on the OpenCode core (also terminal_agent, cloud_agent). | 0 of 4 points · 2 matchups | See pricing |
| 28 | [Kimi Code](https://www.versusref.com/coding-agents/tools/kimi-code/) (OSS) | Open-weight model vendor's agent, included with Kimi membership; MIT CLI also takes other providers. | 0 of 2 points · 1 matchup | See pricing |
| 29 | [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. | 0 of 4 points · 2 matchups | See pricing |
| 30 | [Letta Code](https://www.versusref.com/coding-agents/tools/letta-code/) (OSS) | Persistent-memory coding agent (Apache-2.0); CLI + desktop + browser + Slack/Telegram/Discord channels; optional Letta Cloud ($20/mo Pro). | 0 of 2 points · 1 matchup | See pricing |
| 31 | [Mistral Vibe](https://www.versusref.com/coding-agents/tools/mistral-vibe/) (OSS) | European vendor agent with open-weight local models and self-hosting; also IDE extension and cloud_agent (remote agents, teleport). | 0 of 2 points · 1 matchup | See pricing |
| 32 | [OpenHands](https://www.versusref.com/coding-agents/tools/openhands/) (OSS) | MIT-licensed, model-agnostic agent platform; self-host or use OpenHands Cloud (free individual SaaS, enterprise VPC). Also a cloud/async agent. | 0 of 2 points · 1 matchup | See pricing |
| 33 | [Pi](https://www.versusref.com/coding-agents/tools/pi/) (OSS) | Minimal-by-design harness: no built-in subagents, plan mode or permission popups; extend via packages, skills and SDK. | 0 of 2 points · 1 matchup | See pricing |
| 34 | [Qwen Code](https://www.versusref.com/coding-agents/tools/qwen-code/) (OSS) | Open-source multi-provider agent from an open-weight model lab; also desktop app, web UI and VS Code/JetBrains/Zed plugins. | 0 of 2 points · 1 matchup | See pricing |
| 35 | [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 of 2 points · 1 matchup | See pricing |
| 36 | [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 of 2 points · 1 matchup | See pricing |
| 37 | [Warp](https://www.versusref.com/coding-agents/tools/warp/) | Agentic terminal / ADE with credit-metered agents and cloud agent orchestration ("Warp Factories"). Secondary segment: cloud_agent. Client source available under AGPL-3.0. | 0 of 2 points · 1 matchup | See pricing |
| 38 | [Zed](https://www.versusref.com/coding-agents/tools/zed/) (OSS) | Fast OSS editor with agentic features; free with your own keys or external agents, Pro $10/mo adds $5 of hosted-model tokens at list +10%. Also oss_self_hosted-adjacent (GPL editor, local models). | 0 of 2 points · 1 matchup | See pricing |

### 1. GitLab Duo Agent Platform

Training policy and self-hosting matter most to a security review. [Full GitLab Duo Agent Platform vs GitHub Copilot verdict](https://www.versusref.com/coding-agents/github-copilot-vs-gitlab-duo-agent-platform/)

### 2. Tabnine

Training on company code is the deciding factor for enterprise buyers. [Full Tabnine vs GitHub Copilot verdict](https://www.versusref.com/coding-agents/github-copilot-vs-tabnine/)

### 3. Cursor

Code training policy matters most here. [Full Cursor vs Claude Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-cursor/)

What matters most to a security review is whether company code trains models and whether the agent can run in your own environment. [Full Cursor vs GitHub Copilot verdict](https://www.versusref.com/coding-agents/cursor-vs-github-copilot/)

Both tools clear the basic security checklist, each offering a self-hosted or VPC deployment, a zero data retention option, SOC 2 and admin audit logs. [Full Cursor vs Devin Desktop (formerly Windsurf) verdict](https://www.versusref.com/coding-agents/cursor-vs-devin-desktop/)

What matters most to a security review is whether company code trains models. [Full Cursor vs OpenAI Codex verdict](https://www.versusref.com/coding-agents/cursor-vs-openai-codex/)

For engineering orgs, code handling and deployment control decide this. [Full Cursor vs Google Antigravity verdict](https://www.versusref.com/coding-agents/cursor-vs-google-antigravity/)

Cursor covers more of an enterprise security review. [Full Cursor vs Kiro verdict](https://www.versusref.com/coding-agents/cursor-vs-kiro/)

Both tools address training on company code: Augment Code states no training on paid plans, and Cursor says code is not used for training with Privacy Mode on. [Full Cursor vs Augment Code verdict](https://www.versusref.com/coding-agents/augment-code-vs-cursor/)

Both vendors limit training: Cursor says code is not used for training with Privacy Mode on, and Zed does not train on Zed-hosted model traffic. [Full Cursor vs Zed verdict](https://www.versusref.com/coding-agents/cursor-vs-zed/)

For a security review, Cursor publishes the commitments buyers ask for. [Full Cursor vs Trae verdict](https://www.versusref.com/coding-agents/cursor-vs-trae/)

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

Both make a training commitment: Cursor states that with Privacy Mode on, code is not used for training, and Verdent states that your input and output will not be used for model training. [Full Cursor vs Verdent verdict](https://www.versusref.com/coding-agents/cursor-vs-verdent/)

### 4. Claude Code

Claude Code is the only one here with an enterprise record. [Full Claude Code vs Kimi Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-kimi-code/)

Claude Code covers more of an enterprise security checklist. [Full Claude Code vs GitHub Copilot verdict](https://www.versusref.com/coding-agents/claude-code-vs-github-copilot/)

Neither tool publishes a blanket no-training commitment for company code: Claude Code's consumer plans leave training to user choice, and Google Antigravity's Individual tier is governed by Google Terms of Service. [Full Claude Code vs Google Antigravity verdict](https://www.versusref.com/coding-agents/claude-code-vs-google-antigravity/)

The deciding factor here is training on company code. [Full Claude Code vs Devin verdict](https://www.versusref.com/coding-agents/claude-code-vs-devin/)

Neither publishes a clean no-training guarantee for company code. [Full Claude Code vs Warp verdict](https://www.versusref.com/coding-agents/claude-code-vs-warp/)

On training, the two are comparable: Claude Code leaves it as a user choice on consumer plans, and Mistral Vibe offers an opt-out of model training on all plans. [Full Claude Code vs Mistral Vibe verdict](https://www.versusref.com/coding-agents/claude-code-vs-mistral-vibe/)

OpenCode makes a strong statement on code handling: it does not store any of your code or context data, while Claude Code describes user choice on consumer plans. [Full Claude Code vs OpenCode verdict](https://www.versusref.com/coding-agents/claude-code-vs-opencode/)

Both tools offer self-hosted or VPC deployment. [Full Claude Code vs Aider verdict](https://www.versusref.com/coding-agents/aider-vs-claude-code/)

Claude Code publishes the controls a security review looks for, and Goose publishes none of them. [Full Claude Code vs Goose verdict](https://www.versusref.com/coding-agents/claude-code-vs-goose/)

Claude Code publishes a full enterprise record and Qwen Code publishes none of it. [Full Claude Code vs Qwen Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-qwen-code/)

Claude Code publishes the enterprise controls a security review asks for, and Pi publishes none of them. [Full Claude Code vs Pi verdict](https://www.versusref.com/coding-agents/claude-code-vs-pi/)

Claude Code clears more of an enterprise security review. [Full Claude Code vs Grok Build verdict](https://www.versusref.com/coding-agents/claude-code-vs-grok-build/)

Command Code has the firmer training promise: it states it does not train AI models on your source code, while Claude Code leaves training as a user choice on its consumer plans. [Full Claude Code vs Command Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-command-code/)

Claude Code covers the enterprise checklist, and Letta Code publishes little of it. [Full Claude Code vs Letta Code verdict](https://www.versusref.com/coding-agents/claude-code-vs-letta-code/)

Claude Code covers the enterprise checklist, and DeepSeek Harness (dsh) publishes none of it. [Full Claude Code vs DeepSeek Harness (dsh) verdict](https://www.versusref.com/coding-agents/claude-code-vs-deepseek-harness/)

Both tools offer SOC 2, admin audit logs, SSO and a self-hosted or VPC deployment, so the deciding points are how code is handled and how firmly each claim is backed. [Full Claude Code vs OpenAI Codex verdict](https://www.versusref.com/coding-agents/claude-code-vs-openai-codex/)

Freebuff states that it does not use prompt and project data to train its models, while Claude Code describes user choice on consumer plans, so training is a point in Freebuff's favor. [Full Claude Code vs Freebuff verdict](https://www.versusref.com/coding-agents/claude-code-vs-freebuff/)

### 5. Devin Desktop (formerly Windsurf)

Neither vendor's training policy, as published, settles the question on its own, so the deployment and control features decide it. [Full Devin Desktop (formerly Windsurf) vs GitHub Copilot verdict](https://www.versusref.com/coding-agents/devin-desktop-vs-github-copilot/)

Neither vendor's training policy, as published, decides this on its own, so deployment and data controls carry the comparison. [Full Devin Desktop (formerly Windsurf) vs Kiro verdict](https://www.versusref.com/coding-agents/devin-desktop-vs-kiro/)

### 6. OpenAI Codex

Neither vendor publishes a crisp training statement: GitHub Copilot's policy note reads "From 2026-04-24" and OpenAI Codex's reads "Business". [Full OpenAI Codex vs GitHub Copilot verdict](https://www.versusref.com/coding-agents/github-copilot-vs-openai-codex/)

OpenAI Codex has the fuller enterprise package. [Full OpenAI Codex vs Aider verdict](https://www.versusref.com/coding-agents/aider-vs-openai-codex/)

### 7. OpenCode

Training on company code and self-hosting matter most here. [Full OpenCode vs Kilo Code verdict](https://www.versusref.com/coding-agents/kilo-code-vs-opencode/)

Code handling and deployment control decide this, and only OpenCode publishes anything here. [Full OpenCode vs Crush verdict](https://www.versusref.com/coding-agents/crush-vs-opencode/)

### 8. Cline

A security review starts with code training and deployment control, and Cline answers both. [Full Cline vs Kilo Code verdict](https://www.versusref.com/coding-agents/cline-vs-kilo-code/)

### 9. Amp

What matters most here is whether the vendor trains on company code. [Full Amp vs Claude Code verdict](https://www.versusref.com/coding-agents/amp-vs-claude-code/)

### 10. Factory

For a security review, training on company code matters most. [Full Factory vs Devin verdict](https://www.versusref.com/coding-agents/devin-vs-factory/)

### 11. Gemini CLI

For a security review, the training policy matters most. [Full Gemini CLI vs Google Antigravity verdict](https://www.versusref.com/coding-agents/gemini-cli-vs-google-antigravity/)

### 12. JetBrains Junie

Training policy matters most to a security review, and JetBrains Junie gives the clearer answer: no training on inputs or outputs unless expressly agreed. [Full JetBrains Junie vs GitHub Copilot verdict](https://www.versusref.com/coding-agents/github-copilot-vs-junie/)

### 13. Tabby

Neither tool publishes a clear training commitment: GitHub Copilot's policy note reads "From 2026-04-24" and Tabby publishes none. [Full Tabby vs GitHub Copilot verdict](https://www.versusref.com/coding-agents/github-copilot-vs-tabby/)

### 14. Aider

Aider publishes the more reassuring enterprise posture. [Full Aider vs gptme verdict](https://www.versusref.com/coding-agents/aider-vs-gptme/)

### 15. Devin

Training on company code is the first question, and neither gives a clean answer: Devin's self-serve plans may train by default, and OpenHands publishes no training policy. [Full Devin vs OpenHands verdict](https://www.versusref.com/coding-agents/devin-vs-openhands/)

Devin has the stronger compliance record, with a caveat on training: its self-serve plans may train on code by default, and Cosine publishes no training policy at all. [Full Devin vs Cosine verdict](https://www.versusref.com/coding-agents/cosine-vs-devin/)

### 16. 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.

### 17. Command Code

Coding agent for open models (DeepSeek, Kimi, GLM, Qwen, MiniMax) with taste learning; CLI + desktop + ACP for Zed; cheap $1-$20 plans with published $ rolling windows. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 18. 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.

### 19. 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.

### 20. DeepSeek Harness (dsh)

'Everything is a Plugin' (Cordis) agent harness from DeepSeek; developer preview with breaking changes; general-purpose, not coding-only. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 21. Freebuff

Ad-funded free coding agent on curated low-cost models. History: launched as Codebuff (YC); repo and operator (Freebuff, Inc., legal docs effective 2026-09-02) now present Freebuff, built on the Codebuff open multi-agent framework; codebuff.com still sells the premium-model Codebuff CLI ($100-$500/mo). Also desktop, web builder, cloud IDE. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 22. GitHub Copilot

Multi-model IDE extension plus Copilot cloud agent (formerly 'coding agent'; cloud_agent) and Copilot CLI (terminal_agent). Billing moved from premium requests to GitHub AI Credits on 2026-06-01. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 23. Google Antigravity

Agent-first IDE + CLI + desktop agent manager; Google's consumer coding product since Gemini CLI's 2026-06-18 consumer cutover. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 24. Goose

Local, BYO-model general-purpose agent; also a native desktop app (Rust). Donated by Block to the Linux Foundation's Agentic AI Foundation in 2026. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 25. 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.

### 26. Grok Build

Subscription-bundled vendor terminal agent (SuperGrok / X Premium+), open-sourced Apache-2.0; also embeds in editors via ACP. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 27. Kilo Code

OSS multi-surface agent; zero-markup inference gateway + Kilo Pass credit bundles; started as a Roo Code fork, now built on the OpenCode core (also terminal_agent, cloud_agent). No won verdicts for this use case yet; it ranks on ties and near-misses.

### 28. Kimi Code

Open-weight model vendor's agent, included with Kimi membership; MIT CLI also takes other providers. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 29. 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.

### 30. Letta Code

Persistent-memory coding agent (Apache-2.0); CLI + desktop + browser + Slack/Telegram/Discord channels; optional Letta Cloud ($20/mo Pro). No won verdicts for this use case yet; it ranks on ties and near-misses.

### 31. Mistral Vibe

European vendor agent with open-weight local models and self-hosting; also IDE extension and cloud_agent (remote agents, teleport). No won verdicts for this use case yet; it ranks on ties and near-misses.

### 32. OpenHands

MIT-licensed, model-agnostic agent platform; self-host or use OpenHands Cloud (free individual SaaS, enterprise VPC). Also a cloud/async agent. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 33. Pi

Minimal-by-design harness: no built-in subagents, plan mode or permission popups; extend via packages, skills and SDK. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 34. Qwen Code

Open-source multi-provider agent from an open-weight model lab; also desktop app, web UI and VS Code/JetBrains/Zed plugins. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 35. 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.

### 36. 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.

### 37. Warp

Agentic terminal / ADE with credit-metered agents and cloud agent orchestration ("Warp Factories"). Secondary segment: cloud_agent. Client source available under AGPL-3.0. No won verdicts for this use case yet; it ranks on ties and near-misses.

### 38. Zed

Fast OSS editor with agentic features; free with your own keys or external agents, Pro $10/mo adds $5 of hosted-model tokens at list +10%. Also oss_self_hosted-adjacent (GPL editor, local models). No won verdicts for this use case yet; it ranks on ties and near-misses.

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