Self-hosted AI
coding agent platform
Chat with any LLM. Run coding agents in Docker. Orchestrate multi-role pipelines. Memory that learns. All on your own server.
Chat with Any LLM
Ollama, Anthropic, OpenAI, Google — one interface, any model. Self-hosted or cloud.
Docker Coding Agents
Aider, OpenCode, Codex CLI, Gemini CLI — ephemeral containers that clone, edit, and push.
Memory & RAG
Auto-summarize chats, embed with nomic-embed-text, inject relevant context into every prompt.
Cluster Pipelines
Chain multi-role agents — PR reviewer, docs writer, test coverage — sequentially or in parallel.
Project Connect
Mount local folders, attach files to chat, auto-inject CLAUDE.md into system prompts.
Notifications
Telegram and Slack alerts for agent jobs. GitHub PR comment triggers with /ghostbot.
GitHub Integration
Agents push commits, open PRs, post status. Webhook-driven — comment on a PR to trigger a job.
Self-Hosted & Secure
Your data stays on your server. AES-256-GCM encrypted secrets, JWT auth, multi-user invites.
How it works
One Next.js app. No microservices, no message queues. SQLite is the only datastore.
Built with
Ready to self-host?
Clone the repo, run one command, and you have your own AI coding agent platform.