Ship an AI chat app your whole organization can use.

Turn git-versioned configuration into a live, streaming Claude app for your organization, running on its own Anthropic key.

Start with your agent

Copy one line. The guide handles organization setup, a sample app, and the human-only sign-in and secret steps.

Setup an app with ai.frde.me/agents.md

For browser-capable agents. You approve sign-in and enter secrets yourself.

1
Build
Author the agent in your repo: agent.md for the system prompt, config.json for model and effort, optional theme, tools, and a custom app UI.
2
Release
Cut an immutable, numbered release from any ref. Every release records the commit it came from.
3
Deploy
Point environments at releases. Pushes auto-deploy tracked branches; pull requests get their own preview environments.
4
Verify
Open the app, or let the agent validate its work, open a pull request, and check its own preview build.

Git as source of truth: Connect a repository once per organization. Webhooks build releases on push and previews on pull requests; the dashboard draft covers quick starts.
Streaming chat apps: Token streaming, collapsible work summaries, and live tool-call cards, with conversation history per user and environment.
Agent tools: Opt-in grants let a deployed agent read its repo, propose pull requests, run a sandboxed Linux workspace, and deploy releases.
Bring your own key: Each organization chats on its own Anthropic key, validated before storage and encrypted at rest with AES-256-GCM.
Environments: Production by default, plus branch-tracking and PR preview environments, each with its own stable URL.
Multi-tenant: WorkOS organizations scope every project, key, and connection; switching tenants switches everything.