Prx

OpenPRX

Self-hosted AI assistant framework built in Rust. Multi-channel, multi-provider, with built-in self-evolution.

Forked from ZeroClaw and extended with production reliability, governance-aware AI, and a self-evolution system.

Highlights

• 9 LLM providers — Anthropic, OpenAI, Google Gemini, GitHub Copilot, Ollama, AWS Bedrock, GLM, OpenAI Codex, and OpenAI-compatible endpoints
• LLM Router — heuristic routing (capability + Elo + cost + latency), KNN semantic routing (cold-start guard + 100ms timeout fallback), and Automix low-confidence auto-upgrade
• Causal Tree Engine — speculative multi-branch prediction with rehearsal, scoring, and circuit breaker; opt-in via causal_tree.enabled (disabled by default)
• 19 messaging channels — Signal, WhatsApp, Telegram, Discord, Slack, Matrix, and more
• 38+ built-in tools — shell, MCP, memory, scheduling, remote nodes
• Xin (心) task engine — autonomous heartbeat scheduler with 3 execution modes (Rust/LLM/Shell), 5 built-in system tasks, SQLite persistence
• Web Console — browser-based management interface (console/)
• Remote Nodes — control macOS/Linux/Pi devices via prx-node agent
• Self-Evolution — autonomous prompt/memory/strategy improvement with xin-managed scheduling
• Subagent Governance — concurrency limits, depth control, config inheritance
• 3,500+ tests — comprehensive test coverage across all modules

LLM Router Flags

• heuristic model routing is always available
• semantic KNN scoring activates when embeddings and history are available
• cheap-first, low-confidence upgrade activates when a premium model is configured

Causal Tree Engine Flags

> Disabled by default. Set causal_tree.enabled = true to activate.

• causal_tree.enabled — master switch for the CTE pipeline (default: false)
• causal_tree.policy.max_branches — maximum candidate branches to expand (default: 3)
• causal_tree.policy.commit_threshold — minimum score to commit a branch (default: 0.62)
• causal_tree.policy.extra_latency_budget_ms — max additional latency budget in ms (default: 300)
• causal_tree.policy.rehearsal_timeout_ms — per-rehearsal timeout in ms (default: 5000)
• causal_tree.policy.circuit_breaker_threshold — consecutive failures before tripping (default: 5)
• causal_tree.w_confidence — scoring weight for confidence dimension (default: 0.50)
• causal_tree.w_cost — scoring weight for cost penalty (default: 0.25)
• causal_tree.w_latency — scoring weight for latency penalty (default: 0.25)

Quick Start

# Build
git clone https://github.com/openprx/prx.git && cd prx
cargo build --release --all-features

# Setup
cp target/release/openprx /usr/local/bin/
openprx onboard

# Run
openprx start

Default build (cargo build) includes llm-router.

Or download pre-built binaries from Releases.

Binaries

| Binary | Description | |--------|-------------| | openprx | Main AI daemon — providers, channels, tools, evolution | | prx-node | Lightweight remote node agent — runs on managed devices |

Architecture

         Channels (19)          Tools (38+)           Remote Nodes
    Signal · WA · TG · ...    Shell · MCP · ...     macOS · Pi · ...
              │                      │                     │
              ▼                      ▼                     ▼
         ┌─────────────────────────────────────────────────────┐
         │                    openprx daemon                    │
         │  Agent Loop · Gateway · CTE · Xin · Memory · Evo   │
         └──────────────────────┬──────────────────────────────┘
                                │
                     Providers (9 LLMs)
              Anthropic · OpenAI · Google · ...

Documentation

| Topic | Description | |-------|-------------| | [Providers](docs/providers.md) | 9 LLM providers, fallback chains, token refresh | | [Channels](docs/channels.md) | 19 messaging platforms, DM/group policies | | [Tools](docs/tools.md) | 38 built-in tools, hooks system, webhooks | | [Remote Nodes](docs/remote-nodes.md) | prx-node agent, device pairing, JSON-RPC | | [Web Console](docs/web-console.md) | Browser-based management interface | | [Evolution](docs/evolution.md) | Self-improvement pipeline | | [Configuration](docs/configuration.md) | Config reference, workspace files, security | | [Router](docs/router.md) | LLM Router config, flow, safety boundaries | | [Causal Tree Engine](docs/causal-tree.md) | CTE pipeline, branch prediction, rehearsal, scoring | | [WASM Plugins](docs/plugin-developer-guide.md) | Plugin developer guide (Rust/Python/JS/Go) | | [Host Function Reference](docs/host-function-reference.md) | WASM plugin host API reference | | [Plugin Runtime Lifecycle](docs/plugin-runtime-lifecycle.md) | Atomic generations, subscriber pumps, trust and hook bounds |

Links

• Documentation — Full PRX documentation (10 languages)
• Community — OpenPRX community forum
• OpenPRX — Project homepage

Related Projects

| Repository | Description | |------------|-------------| | openprx/prx | AI assistant framework (this repo) | | openprx/prx-memory | Standalone memory MCP server | | openprx/openpr | Project management platform | | openprx/openpr-webhook | Webhook receiver for OpenPR | | openprx/wacli | WhatsApp CLI with JSON-RPC daemon |

Origin & License

Forked from zeroclaw-labs/zeroclaw (MIT / Apache-2.0). "ZeroClaw" is a trademark of ZeroClaw Labs. This project is OpenPRX, an independent fork.

Dual-licensed under [MIT](LICENSE-MIT) and [Apache-2.0](LICENSE-APACHE).

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: openprx
• Source: openprx/prx
• License: Apache-2.0
• Homepage: https://openprx.dev/prx

Install and usage instructions live in the source repository linked above.