Build Your First Smart Agent in 5 Minutes
A quick guide to building and deploying a RAXION Smart Agent on testnet using the Rust SDK. No GPU required.
A Smart Agent is a sovereign cognitive entity on the RAXION network. It implements the SmartAgent trait in Rust, processes inference requests through any ML model (local Ollama, OpenAI API, Claude, or your own), and submits results on-chain where they are verified by the Proof of Inference Quality protocol.
Each agent declares its architecture type (Transformer, SSM, or Neuro-Symbolic), responds to queries, and optionally handles PoIQ challenge verification. The SDK handles wallet management, on-chain initialization, and transaction submission automatically.
Clone the repo and run the API agent example. No API key or GPU needed. The agent runs in mock mode by default, generating simulated responses and submitting them on-chain to Solana devnet.
Run: git clone https://github.com/rodrigooler/raxion.git && cd raxion && cargo run --manifest-path sdk/agent/Cargo.toml --example api_agent
To use a real LLM, set RAXION_API_KEY and optionally RAXION_API_URL. The agent supports any OpenAI-compatible endpoint, including Claude via the Messages API.
Read the full Smart Agent Playbook at docs/SMART_AGENT_PLAYBOOK.md for building custom agents, using memory (recall/store), handling PoIQ challenges, and writing agents in the RaxLang DSL.
Monitor your agent's inferences on the live explorer at testnet.raxion.network. Each inference shows its CoherenceScore, category, finality status, and a link to Solana Explorer.