Getting Started

You'll be up and running in under 5 minutes. No API keys required for the demo backend.

Prerequisites

Rust 1.80+ — install via rustup
An LLM API key (Claude, OpenAI, or Gemini) for live runs — OR use the demos which need no keys

1. Install from Source

git clone https://github.com/fboiero/Agentor.git
cd Agentor
cargo build --release

The binary lands at target/release/argentor.

2. Run the Demo (No API Keys)

The fastest way to see Argentor in action:

cargo run -p argentor-cli --example demo_full_pipeline

This runs an 8-step pipeline with real tool execution: shell commands, file I/O, vector memory, and report generation. No mocks, no API keys.

More demos

# DevOps team simulation (4 specialized agents)
cargo run -p argentor-cli --example demo_team

# Skills toolkit showcase (18 utility skills)
cargo run -p argentor-cli --example demo_skills_toolkit

# Multi-agent SaaS factory
cargo run -p argentor-cli --example demo_saas_factory

# Security challenge (penetration testing agent)
cargo run -p argentor-cli --example demo_security_challenge

3. Start the Gateway

# Set your LLM provider key
export ANTHROPIC_API_KEY="sk-ant-..."
# Or: export OPENAI_API_KEY="sk-..."

# Start the server
cargo run -p argentor-cli -- serve --bind 0.0.0.0:8080

Open your browser:

URL	What it is
`http://localhost:8080/dashboard`	Control plane dashboard
`http://localhost:8080/playground`	Interactive chat playground
`http://localhost:8080/health`	Health check
`http://localhost:8080/openapi.json`	OpenAPI 3.0 specification
`http://localhost:8080/metrics`	Prometheus metrics

4. Chat via REST API

Synchronous

curl -X POST http://localhost:8080/api/v1/agent/chat \
  -H "Content-Type: application/json" \
  -d '{"message": "What files are in the current directory?"}'

Streaming (SSE)

curl -N -X POST http://localhost:8080/api/v1/chat/stream \
  -H "Content-Type: application/json" \
  -d '{"message": "Write a haiku about Rust"}'

5. Python SDK

pip install argentor-sdk

from argentor import ArgentorClient

client = ArgentorClient("http://localhost:8080")

# Simple chat
response = client.run_task("Summarize the README.md file")
print(response.output)

# List available skills
skills = client.list_skills()
for skill in skills:
    print(f"  {skill.name}: {skill.description}")

6. Rust Library

[dependencies]
argentor-core    = "1.0"
argentor-agent   = "1.0"
argentor-skills  = "1.0"
argentor-builtins = "1.0"

use argentor_agent::{AgentRunner, ModelConfig, LlmProvider};
use argentor_skills::SkillRegistry;
use argentor_builtins::register_builtins;

#[tokio::main]
async fn main() -> anyhow::Result<()> {
    let mut registry = SkillRegistry::new();
    register_builtins(&mut registry);

    let config = ModelConfig::new(LlmProvider::Claude)
        .with_model("claude-sonnet-4-20250514")
        .with_max_tokens(4096);

    let mut runner = AgentRunner::new(config, registry);
    let response = runner.run("What tools do you have available?").await?;
    println!("{}", response.content);
    Ok(())
}

With guardrails

use argentor_agent::GuardrailEngine;

let guardrails = GuardrailEngine::default(); // PII + injection + toxicity
let mut runner = AgentRunner::new(config, registry)
    .with_guardrails(guardrails)
    .with_cache(1000, std::time::Duration::from_secs(300));

let response = runner.run("Process this customer data").await?;

7. Docker

docker run -d \
  --name argentor \
  -p 8080:8080 \
  -e ANTHROPIC_API_KEY="sk-ant-..." \
  ghcr.io/fboiero/argentor:latest serve

Key Concepts

Skills

Skills are tools the agent can use. Argentor ships 50+ built-in skills (files, shell, git, web search, crypto, etc.) plus a WASM plugin system for custom skills.

cargo run -p argentor-cli -- skill list

Guardrails

Pre/post-execution filters that scan for PII, prompt injection, and policy violations. Integrated directly into the agent loop — zero configuration required.

Multi-Agent Orchestration

use argentor_orchestrator::{Orchestrator, OrchestratorConfig};

let config = OrchestratorConfig::default();
let orchestrator = Orchestrator::new(config);
orchestrator.run_pipeline("Build a REST API for a todo app").await?;

MCP Integration

Argentor acts as both MCP client (connect to MCP servers) and MCP server (expose skills as MCP tools).

cargo run -p argentor-cli -- mcp serve

What's Next?

Benchmarks — how Argentor compares to LangChain, CrewAI, PydanticAI
Enterprise — RBAC, multi-tenancy, readiness endpoint
Deployment Guide — Docker, Kubernetes, Helm
docs.rs — full API reference