Installation

Requirements

• Rust 1.75+ (MSRV — Minimum Supported Rust Version)
• Tokio runtime (TraitClaw is async-first)

Quick Install

cargo add traitclaw traitclaw-openai-compat tokio anyhow

This gives you the core framework with the OpenAI-compatible provider (works with OpenAI, Ollama, Groq, Mistral, and vLLM).

Feature Flags

The traitclaw meta-crate re-exports everything through feature flags:

[dependencies]
# Minimal — just core + macros + OpenAI-compatible provider
traitclaw = "1.0"

# With specific features
traitclaw = { version = "1.0", features = ["steering", "sqlite", "mcp"] }

# Everything
traitclaw = { version = "1.0", features = ["full"] }

Available Features

Feature	Crate	Default	Description
openai-compat	traitclaw-openai-compat	✅	OpenAI, Ollama, Groq, Mistral, vLLM
macros	traitclaw-macros	✅	#[derive(Tool)] proc macro
steering	traitclaw-steering	❌	Guards, Hints, Trackers
sqlite	traitclaw-memory-sqlite	❌	SQLite persistent memory
mcp	traitclaw-mcp	❌	Model Context Protocol client
rag	traitclaw-rag	❌	RAG pipeline (BM25, embeddings)
team	traitclaw-team	❌	Multi-agent orchestration
eval	traitclaw-eval	❌	Evaluation & benchmarking
strategies	traitclaw-strategies	❌	ReAct, CoT, MCTS reasoning
full	All of the above	❌	Everything

Provider Setup

OpenAI

export OPENAI_API_KEY="sk-..."

use traitclaw_openai_compat::OpenAiCompatProvider;

let provider = OpenAiCompatProvider::openai("gpt-4o-mini", std::env::var("OPENAI_API_KEY")?);

Anthropic

cargo add traitclaw-anthropic
export ANTHROPIC_API_KEY="sk-ant-..."

use traitclaw_anthropic::AnthropicProvider;

let provider = AnthropicProvider::new("claude-sonnet-4-20250514", std::env::var("ANTHROPIC_API_KEY")?);

Ollama (Local)

No API key needed — just run Ollama locally:

ollama pull llama3.2

use traitclaw_openai_compat::OpenAiCompatProvider;

let provider = OpenAiCompatProvider::ollama("llama3.2");

Groq / Mistral / vLLM

// Groq
let provider = OpenAiCompatProvider::groq("llama-3.3-70b-versatile", groq_key);

// Mistral
let provider = OpenAiCompatProvider::new("https://api.mistral.ai/v1", "mistral-large", key);

// Any OpenAI-compatible endpoint
let provider = OpenAiCompatProvider::new("http://localhost:8000/v1", "my-model", "no-key");

Next Steps

Quick Start → Build your first agent in 5 minutes.