# Trebell Code > Trebell Code is Trebell's flagship product: a free desktop workspace for AI coding agents. The current release is 2.1.6, for Windows 10 and 11 (x64). ## Company Trebell is a founder-led developer-tools startup founded in May 2026. - Company: Trebell - Company website: https://trebell.dev/ - Founding team and leadership: https://trebell.dev/about/ - Founder & CEO: Tanishq Baweja - Current core team size: 5 - Core team: Tanishq Baweja (Founder & CEO), Raghav Dabra (Engineering), Himnish Ishaan (Product & Research), Shashwat Chaudhary (Operations & Growth), Samriddhi Choursarisa (Engineering) - Business contact: founder@trebell.dev - Founder LinkedIn: https://www.linkedin.com/in/tanishq-baweja-427468191 - Founder GitHub: https://github.com/tanishqbaweja - Company trust center: https://trebell.dev/trust/ - Security and responsible disclosure: https://trebell.dev/security/ - Privacy policy: https://trebell.dev/privacy/ - Terms of use: https://trebell.dev/terms/ ## Product - Trebell Code: https://code.trebell.dev/ - Current release: 2.1.6, Windows 10 and 11 (x64) installer - Download (starts the installer download): https://github.com/tanishqbaweja/trebellcode-releases/releases/download/v2.1.6/Trebell-Code-Setup-2.1.6.exe - Release notes: https://github.com/tanishqbaweja/trebellcode-releases/releases/tag/v2.1.6 - Price: free to download and use - Licensing and releases: https://trebell.dev/licensing/ - Updates: from 2.0.0 on, the app checks the public releases repository for new versions; an update downloads and installs only when the user chooses. ## What it does Trebell Code is a desktop coding workspace around coding agents: files, terminal, Git and pull requests, worktrees, previews, a browser panel, projects and environments. Every turn keeps an activity card that shows what it did, such as a command's exit code and duration or an edit's diff. ## Harnesses - Trebell Native: Trebell's own agent harness, with its own model and tool loop calling model APIs directly. OpenAI is the default provider. - Codex: runs the Codex CLI the user has installed. - Claude Code - OpenCode: OpenCode 1.x and OpenCode 2. - Cursor - Grok Build - Antigravity: runs with a Gemini API key from Google AI Studio. - Muse Code: Meta's coding agent, new in 2.0; sign in with `muse login`. Each of the seven external agents runs as its real harness, with its own models, sign-in, approvals and updates. ## What's new in 2.0 - Muse Code is a new harness: its replies, reasoning, commands and file edits stream into the thread, its approvals and questions come to Trebell's own card and dialog, and Usage shows its 5-hour and weekly windows. - OpenCode 2 is supported beside OpenCode 1.x. - Each harness brings its own model list, options and default; a reopened thread resumes on the model it used; a harness's thinking shows as a collapsed Thinking row. - A redesigned interface: one design across the app, the same top bar on every page, and replies as GitHub-style Markdown with tables and copyable code blocks. - Every turn shows what it did, and a turn that failed or was stopped before it replied says how and why. - Trebell Native streams replies on OpenAI (code-change turns are checked before the answer), can look at images in the workspace, and waits out network drops instead of failing the turn. - Harness updates from Settings. - Lighter on the machine: fewer background processes, less CPU and GPU while typing and streaming, and large threads open with much less disk and database work. ## Intended users Trebell Code is built for software developers and engineering teams using AI coding agents on real codebases, especially where provider choice, reproducible execution, verification, and cost visibility matter. ## Business model Trebell Code is free for developers. Trebell plans a paid team infrastructure layer with shared policy, usage and cost controls, managed execution environments, auditability, and administration. ## Benchmarks Trebell Native was compared with Codex using the OpenAI API, with the same model and settings on both: gpt-6-luna, max reasoning, fast service tier, hosted web search off. Tasks were fresh (never used before), and an independent verifier graded each run. Cost is API-equivalent, compared on the tasks where both got the same verifier result. - SWE-bench Multilingual, 156 fresh tasks: Trebell Native resolved 125, Codex using the OpenAI API resolved 124. Native's cost was 0.819x (95% bootstrap interval 0.75-0.90), 18% less. - Terminal-Bench 2.0, 78 fresh tasks: both resolved 62. Native's cost was 0.721x (0.58-0.87), 28% less. - Terminal-Bench Pro 1.0, 40 fresh tasks: Trebell Native resolved 22, Codex using the OpenAI API resolved 21. Cost was statistically level (0.929x, 0.80-1.08). These results compare Trebell Native with Codex using the OpenAI API under the settings above; read them as that comparison, not as a claim about every harness or setup.