main.py
Output
Run your code to see output here
Master the AI agent runtime - tools, sandboxing, workflows, guardrails, and more. Write and run real Python code directly in your browser.
pip install agentu
Create agents, attach tools, execute with type-safe schemas.
READONLY / WRITE / DANGEROUS access control for tools.
SQLite-backed remember/recall with importance-weighted search.
SHA-256 exact match, semantic, offline, and distributed presets.
NoPII, NoHallucination with automatic self-correction.
Sequential, parallel, fan-out/fan-in with operator chaining.
Test suites, validators, accuracy metrics, LLM-as-judge.
Auto-tracked events, metrics, and full execution traces.
Separate read/write tool buckets with least-privilege defaults.
Multi-turn stateful conversations with isolated contexts.
Progressive-load domain expertise. 3-level context optimization.
Route queries through Ollama, OpenAI, or Anthropic.
Discover and use tools from external Model Context Protocol servers.
Serve agents via HTTP, WebSocket, and SSE with one call.
Feedforward AGENTS.md rules prepended to every LLM call.
CostTracker, Logger, Retry - composable processing pipeline.