Skip to content
Back to search

helix

♾️ Private Agent Fleet with Spec Coding. Each agent gets their own GPU-accelerated desktop. Run Claude, Codex, Gemini and open models on a full private AI Stack ♾️

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

It has changed since its scan on 11 Nov 2026.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/helixml/helix with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/helixml/helix'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/helixml--helix/badge.svg)](/listings/helixml--helix)

HTML

<a href="/listings/helixml--helix"><img src="/api/listings/helixml--helix/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Install on Docker

Use our quickstart installer:

curl -sL -O https://get.helixml.tech/install.sh
chmod +x install.sh
sudo ./install.sh

The installer will prompt you before making changes to your system. By default, the dashboard will be available on `http://localhost:8080`.

For setting up a deployment with a DNS name, see `./install.sh --help` or read [the detailed docs](https://helix.ml/docs). We've documented easy TLS termination for you.

**Next steps:** - Attach your own GPU runners per [runners docs](https://helix.ml/docs) - Use any [external OpenAI-compatible LLM](https://helix.ml/docs) (including self-hosted vLLM)

Install on Kubernetes

Use our Helm charts for production deployments: - [Control Plane Helm Chart](https://helix.ml/docs) - [Runner Helm Chart](https://helix.ml/docs)

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 217 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. <div align="center">
  2. <img alt="logo" src="https://helix.ml/assets/helix-logo.png" width="250px">
  3. <br/>
  4. <br/>
  5. </div>
  6. <p align="center">
  7. <a href="https://app.helix.ml/">SaaS</a> •
  8. <a href="https://helix.ml/docs">Private Deployment</a> •
  9. <a href="https://helix.ml/docs">Docs</a> •
  10. <a href="https://discord.gg/VJftd844GE">Discord</a>
  11. </p>
  12. # Helix — a private agent fleet with spec-driven coding
  13. [👥 Discord](https://discord.gg/VJftd844GE)
  14. **Run a fleet of coding agents on your own infrastructure. Each agent gets its own GPU-accelerated desktop; you organize the work on a spec-driven Kanban board and review the pull requests.**
  15. **It's about running agents on the server, not on every developer's laptop.** You already run Claude Code (or Codex, or Gemini) locally: one agent, one terminal, tied to your machine and your attention. You wouldn't hire a team of developers and sit them all at one laptop — so why make your agents share one? Helix gives each agent its own computer.
  16. Helix runs those agents as a **fleet** — many in parallel, each in its own isolated sandbox with a full desktop (browser, terminal, filesystem, GUI apps), on your own infrastructure, shared with your team. You describe an outcome as a **spec task**, an agent plans it, you approve the plan, and it implements the change and opens a pull request. Instead of one person babysitting one agent in one terminal, you and your team collaboratively dispatch, steer, and review many from a shared Kanban board.
  17. It runs with the agent harnesses you already use (Claude Code, Codex, Gemini CLI, Qwen Code, or any ACP-compatible agent) against the LLM providers you already use — including self-hosted models on your own GPUs. Self-hostable end to end, including air-gapped.
  18. ![Helix Kanban board with a spec task being created — this README rewrite was itself one of these tasks](docs/images/kanban-board.png)
  19. <p align="center"><em>Yes, this is meta: the task that produced this README is the "New SpecTask" open on the right.</em></p>
  20. ## Projects & the Kanban board
  21. Work lives in **projects**. Each project connects one or more git repositories and has a Kanban board where **spec tasks** move left to right:
  22. **Backlog → Planning → Spec Review → In Progress → Pull Request → Merged**
  23. 1. **Backlog** — Create a task with a one-paragraph description of the outcome you want (what should be true when done, not how to do it).
  24. 2. **Planning** — Click **Start Planning**. A planning agent reads the repo and writes a spec (requirements, design, task breakdown) to a `helix-specs` branch.
  25. 3. **Spec Review** — Read the plan. Highlight text to request changes and the agent re-plans, or **Approve** to commit to it.
  26. 4. **In Progress** — An implementation agent codes in its own isolated desktop sandbox. Watch it live, type into the task thread to steer it, or switch to a different agent mid-session — the new one picks up where the last left off.
  27. 5. **Pull Request** — When it's done, a PR is opened in your repo. The PR is the real review gate.
  28. 6. **Merged** — The task closes when the PR merges.
  29. Tasks run **in parallel** — each in its own sandbox — so you're not waiting for one to finish before starting the next. Per-column WIP limits keep the board honest. See [Manage your backlog on the Kanban board](https://helix.ml/docs).
  30. ## Why Helix is different
  31. - **A full desktop per agent — not just a terminal.** Every agent gets a GPU-accelerated streaming desktop with a browser, terminal, filesystem, and GUI apps. You can watch any agent work in real time.
  32. - **Fleet visibility.** See every running agent from a dashboard, zoom into any one's live screen, and jump in with pair programming when it gets stuck.
  33. - **Multiplayer.** Agent environments are shared. Teammates across time zones open the same task and keep going — full chat history and running state persist, no handoff summaries.
  34. - **High-density isolation.** Many fully isolated agent desktops run on a single machine, with a deduplicated filesystem and per-agent credential and network isolation.
  35. - **No lock-in, and cheaper runs.** Swap agent harnesses per task and point Helix at whatever models you run. Because context carries across the switch, you can start a task on a cheap local model and escalate to an expensive one only when it's actually needed — instead of paying frontier prices for the whole run.
  36. ## Works with your stack (no lock-in)
  37. **Agent harnesses** — use any of these per task, and swap between them mid-task:
  38. - Claude Code
  39. - OpenAI Codex
  40. - Gemini CLI
  41. - Qwen Code
  42. - Goose
  43. - Zed Agent
  44. - Anything that speaks **ACP (Agent Client Protocol)**
  45. **Source control** — connect your repositories and Helix opens pull requests where your code already lives:
  46. - GitHub
  47. - GitLab
  48. - Azure DevOps
  49. **LLM providers** — hosted or self-hosted:
  50. - Major hosted providers (OpenAI, Anthropic, …)
  51. - **Anthropic via Helix's proxy**, including **Anthropic on Google Vertex AI** and **Anthropic on AWS Bedrock**
  52. - **Self-hosted models** — attach any OpenAI-compatible endpoint as an external provider. This includes **[vLLM](https://github.com/vllm-project/vllm)**: point Helix at your vLLM server's OpenAI-compatible URL and run open models on your own GPUs, on Kubernetes or bare metal, air-gapped if you need to.
  53. ## Also included
  54. Helix is a full private GenAI stack, so the pieces you'd expect are here too:
  55. - **Knowledge / RAG** — document ingestion (PDF, Word, text), a web scraper, multiple RAG backends (Kodit, LlamaIndex), PGVector embeddings, and vision RAG.
  56. - **Skills & tools** — REST/OpenAPI integrations, MCP server compatibility, GPTScript, OAuth token management, and a custom-tool SDK.
  57. - **Tracing & observability** — every agent step, requests/responses to LLMs, APIs, and MCP servers, token usage, and cost analysis.
  58. - **Multi-tenancy & auth** — organizations, teams, and role-based access control, with **OIDC** single sign-on. A first SSO sign-in whose email already belongs to a Helix account is linked to that account only when the identity provider asserts `email_verified`; otherwise it is refused (HTTP 409) rather than creating a second account. Adding an organization or team member by email only resolves accounts that have verified the address (IdP assertion, admin-created, accepted invitation, or completed password reset); other addresses are invited.
  59. - **Billing & metering** — track token usage and spend per user and per team, so you can see where the budget goes.
  60. - **Automation** — scheduled/cron tasks and webhook triggers.
  61. - **Notifications** — Slack, Discord, and email.
  62. <img width="1768" alt="AI agents interface" src="https://github.com/user-attachments/assets/0e945ace-4f54-46a2-8d20-49485169486f" />
  63. ## 🚀 Quick Start
  64. ### Install on Docker
  65. Use our quickstart installer:
  66. ```bash
  67. curl -sL -O https://get.helixml.tech/install.sh
  68. chmod +x install.sh
  69. sudo ./install.sh
  70. ```
  71. The installer will prompt you before making changes to your system. By default, the dashboard will be available on `http://localhost:8080`.
  72. For setting up a deployment with a DNS name, see `./install.sh --help` or read [the detailed docs](https://helix.ml/docs). We've documented easy TLS termination for you.
  73. Existing Docker installs run PostgreSQL 12, which is end-of-life. To move to a supported version, follow [the PostgreSQL upgrade guide](docs/postgres-upgrade.md).
  74. **Next steps:**
  75. - Attach your own GPU runners per [runners docs](https://helix.ml/docs)
  76. - Use any [external OpenAI-compatible LLM](https://helix.ml/docs) (including self-hosted vLLM)
  77. ### Install on Kubernetes
  78. Use our Helm charts for production deployments:
  79. - [Control Plane Helm Chart](https://helix.ml/docs)
  80. - [Runner Helm Chart](https://helix.ml/docs)
  81. ## 🔧 Configuration
  82. All server configuration is done via environment variables. You can find the complete list of configuration options in [`api/pkg/config/config.go`](https://github.com/helixml/helix/blob/main/api/pkg/config/config.go).
  83. **Key environment variables:**
  84. - `OPENAI_API_KEY` - OpenAI API credentials
  85. - `ANTHROPIC_API_KEY` - Anthropic API credentials
  86. - `POSTGRES_*` - Database connection settings
  87. - `SERVER_URL` - Public URL for the deployment
  88. - `RUNNER_*` - GPU runner configuration
  89. See the [configuration documentation](https://helix.ml/docs) for detailed setup instructions.
  90. ## 👨‍💻 Development
  91. For local development, refer to the [Helix local development guide](./local-development.md).
  92. **Prerequisites:**
  93. - Docker Desktop (or Docker + Docker Compose)
  94. - Go 1.24.0+
  95. - Node.js 18+
  96. - Make
  97. **Quick development setup:**
  98. ```bash
  99. # Clone the repository
  100. git clone https://github.com/helixml/helix.git
  101. cd helix
  102. # Start supporting services
  103. docker-compose up -d postgres
  104. # Run the backend
  105. cd api
  106. go run . serve
  107. # Run the frontend (in a new terminal)
  108. cd frontend
  109. npm install
  110. npm run dev
  111. ```
  112. See [`local-development.md`](./local-development.md) for comprehensive setup instructions.
  113. ## 📖 Documentation
  114. - **[Overview](https://helix.ml/docs)** - Platform introduction
  115. - **[Getting Started](https://helix.ml/docs)** - Build your first agent
  116. - **[Manage your backlog on the Kanban board](https://helix.ml/docs)** - Projects and spec tasks
  117. - **[Control Plane Deployment](https://helix.ml/docs)** - Production deployment guide
  118. - **[Runner Deployment](https://helix.ml/docs)** - GPU runner setup
  119. - **[API Reference](https://helix.ml/docs)** - REST API documentation
  120. - **[Contributing Guide](./CONTRIBUTING.md)** - How to contribute
  121. - **[Upgrading Guide](./charts/helix-controlplane/UPGRADE.md)** - Control plane upgrade instructions
  122. ## 🤝 Contributing
  123. We welcome contributions! Please see our [Contributing Guide](./CONTRIBUTING.md) for details.
  124. By contributing, you confirm that:
  125. - Your changes will fall under the same license
  126. - Your changes will be owned by HelixML, Inc.
  127. ## 📄 License
  128. Helix is [licensed](https://github.com/helixml/helix/blob/main/LICENSE.md) under a similar license to Docker Desktop. You can run the source code (in this repo) for free for:
  129. - **Personal Use:** Individuals or people personally experimenting
  130. - **Educational Use:** Schools and universities
  131. - **Small Business Use:** Companies with under $10M annual revenue and less than 250 employees
  132. If you fall outside of these terms, please use the [Launchpad](https://deploy.helix.ml) to purchase a license for large commercial use. Trial licenses are available for experimentation.
  133. You are not allowed to use our code to build a product that competes with us.
  134. ### Why these license clauses?
  135. - We generate revenue to support the development of Helix. We are an independent software company.
  136. - We don't want cloud providers to take our open source code and build a rebranded service on top of it.
  137. If you would like to use some part of this code under a more permissive license, please [get in touch](mailto:info@helix.ml).
  138. ## 🆘 Support
  139. - **[Discord Community](https://discord.gg/VJftd844GE)** - Join our community for help and discussions
  140. - **[GitHub Issues](https://github.com/helixml/helix/issues)** - Report bugs or request features
  141. - **[Documentation](https://helix.ml/docs)** - Comprehensive guides and references
  142. - **[Email](mailto:info@helix.ml)** - Contact us for commercial inquiries
  143. ## 🌟 Star History
  144. If you find Helix useful, please consider giving us a star on GitHub!
  145. ---
  146. Built with ❤️ by [HelixML, Inc.](https://helix.ml)

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.