Overview
OpenPhoenix is a single-binary personal AI agent runtime written in Rust. It combines chat, background service, web UI, and CLI in one static binary that can be carried on a USB stick. The project states that it requires no Node, no Python, no runtime to install, and no telemetry. It works with 21 providers, speaks 3 API dialects, and can go live on 9 chat platforms in a single command. Configuration is stored in one TOML file with mode 600, and credentials come from the host environment, supported OAuth flows, or a sealed local store.
On first run, OpenPhoenix finds existing provider credentials or a local Ollama, writes the config, and picks a model that answers. It offers three memory modes: ghost forgets all, session forgets on restart, and recall keeps plain-text notes the user controls. Fallback chains allow switching model, provider, key, and endpoint mid-conversation, and multiple keys per provider rotate on rate limits. Colab mode lets two models plan, debate, divide the work, execute concurrently, and peer-review the result as equal seats.
Security features include a workspace file jail, command gate, secret scrubbing, SSRF blocking, and session quarantine, with fail-closed defaults. The runtime supports 9 chat channels: Telegram, WhatsApp, Discord, Slack, Signal, iMessage, IRC, Matrix, and Mattermost. It has 7 direct dependencies, no async runtime, and no runtime to install. An MCP client uses real JSON-RPC over stdio, and phoenix doctor provides a diagnostic check for config validity, permissions, service status, and security posture. OpenPhoenix is available for Linux, macOS, and Windows.
Pricing
Free
Detailed plans have not been confirmed in our catalog. Check the official website for current limits and billing terms.
Visit WebsitePrices and limits may change. Confirm the currency, billing period, seat minimum and usage caps on the official website.
Use Cases
- Users use OpenPhoenix to run a personal AI agent from a single static binary without installing Node or Python.
- Users use OpenPhoenix to connect one agent to 21 providers and 9 chat platforms with a single command.
- Users use OpenPhoenix to configure credentials and settings through one TOML file with restricted permissions.
- Users use OpenPhoenix to control memory retention with ghost, session, or recall modes.
- Users use OpenPhoenix to switch models, providers, keys, and endpoints mid-conversation and rotate keys on rate limits.
- Users use OpenPhoenix to run two models in Colab for planning, debate, concurrent execution, and peer review.
- Users use OpenPhoenix to harden agent operations with file jail, command gate, secret scrubbing, SSRF blocking, and session quarantine.
- Users use OpenPhoenix to connect MCP servers as tools over JSON-RPC stdio.
Features
Runs personal AI agents
Ships as a single binary
Supports 21 AI providers
Connects through nine chat channels
Runs without runtime dependencies
Built with the Rust language
Reviews
No reviews yet. Be the first to share your experience!