Skip to content
OpenGPT0
HomeDocs

Organizations

Use OpenGPT0 safely in an organization.

OpenGPT0 is open source. The current model-provider surface is Browser/ChatGPT plus the built-in local providers; it does not include a hosted commercial gateway or sales-only edition. For organization rollouts, configure the built-in provider paths explicitly and keep data handling decisions under your own control.

Provider Strategy

Use the provider that matches your organization’s policy:

  • Browser: use browser/chatgpt when your team is allowed to use ChatGPT through a dedicated authenticated browser profile.
  • Ollama: use ollama/... when you want model execution to stay on local developer machines or organization-managed runners.
  • Rapid MLX: use rapid-mlx/... when you run a separate local MLX OpenAI-compatible endpoint.

You can make this explicit in config:

opengpt0.json
{
  "$schema": "https://opengpt0.com/config.json",
  "model": "browser/chatgpt",
  "experimental": {
    "policies": [
      { "effect": "deny", "action": "provider.use", "resource": "*" },
      { "effect": "allow", "action": "provider.use", "resource": "browser" },
      { "effect": "allow", "action": "provider.use", "resource": "ollama" },
      { "effect": "allow", "action": "provider.use", "resource": "rapid-mlx" }
    ]
  }
}

Data Handling

OpenGPT0 reads local files, runs shell commands, edits files, and performs Git operations through its permissioned tools. The Browser provider is treated as an untrusted model source; it should not directly access local files, shell, Git, browser state, cookies, or secrets.

For the most isolated setup, use a built-in local provider with local models:

Terminal window
ollama pull qwen3.5:9b-mlx
opengpt0.json
{
  "$schema": "https://opengpt0.com/config.json",
  "model": "ollama/qwen3.5:9b-mlx"
}

Sharing

If a user enables /share, the conversation and associated session data are sent to the service that hosts share pages at opengpt0.com. Disable sharing for organization rollouts unless your policy allows it.

opengpt0.json
{
  "$schema": "https://opengpt0.com/config.json",
  "share": "disabled"
}

Learn more about sharing.

Private Registries

OpenGPT0 supports private npm registries through Bun’s native .npmrc support. If your organization uses a private registry, such as JFrog Artifactory or Nexus, ensure developers are authenticated before running OpenGPT0.

Terminal window
npm login --registry=https://your-company.jfrog.io/api/npm/npm-virtual/

This creates ~/.npmrc with authentication details. OpenGPT0 automatically uses the registry settings already available to Bun.

You can also configure .npmrc manually:

~/.npmrc
registry=https://your-company.jfrog.io/api/npm/npm-virtual/
//your-company.jfrog.io/api/npm/npm-virtual/:_authToken=${NPM_AUTH_TOKEN}