elie222/rakazo

Rakazo is an open-source platform for running persistent AI agents that you can deploy on your own infrastructure.

The tool addresses the need to run AI agents without vendor lock-in by letting you bring your own language model and compute provider. It works by providing a unified interface across web, desktop, and mobile clients that connect to a self-hosted backend. Agents persist across sessions with their own conversations, memory, and routines. They can access browsers, terminals, files, and graphical desktops, and can delegate work to peer agents or short-lived subagents. The platform supports multiple compute sandboxes including Docker, E2B, Daytona, and Box, plus local execution. Tool integrations come through Composio, Pipedream Connect, MCP servers, Treg endpoints, and OpenAPI documents.

Rakazo suits teams wanting to self-host AI agents without depending on a single vendor's model or infrastructure. It works well for projects requiring persistent agent state, multi-agent coordination, or integration with existing tools and APIs. The platform is available as a web application, Electron desktop app, or Expo mobile app, so you can choose your deployment surface. Quick-start installation via Docker Compose requires only Docker Engine, curl, and OpenSSL with no source checkout needed. Local development requires Node.js 22 and pnpm 9.

The project shows active development with regular commits across the full stack. Work spans frontend improvements in React and Vite, backend changes in Hono and PostgreSQL, and ongoing integration with external services like Composio and Pipedream. The codebase maintains TypeScript throughout and includes tooling for both published Docker images and source-based deployment. Documentation covers self-hosting, local development setup, and agent-assisted installation flows.