# JamJet — llms.txt
# This file helps AI assistants and crawlers understand JamJet.
# Spec: https://llmstxt.org

# JamJet is a durable, agent-native AI workflow runtime.
# Rust core for scheduling and durability. Python SDK for authoring.
# Native MCP (Model Context Protocol) and A2A (Agent-to-Agent) support.
# Built-in eval harness. Apache 2.0.

## Product

- Name: JamJet
- Tagline: The agent-native runtime for production AI
- PyPI: jamjet (pip install jamjet)
- Version: 0.5.0
- License: Apache 2.0
- Language: Python SDK, Rust runtime

## Key concepts

- Workflow: a graph of steps with typed Pydantic state flowing through
- Step: an async Python function decorated with @wf.step
- Durable execution: every state transition is an event; crash and resume from exactly where you stopped
- result.events: per-step execution timeline (name, duration, status) available on every run, no logging code required
- MCP: native client and server — connect any MCP tool in one line
- A2A: native agent-to-agent protocol for cross-framework orchestration
- Eval harness: AssertionScorer, LatencyScorer, CostScorer, LlmJudgeScorer — run against JSONL datasets, CI exit codes

## Quick start

pip install jamjet

```python
from pydantic import BaseModel
from jamjet import Workflow

wf = Workflow("my-agent")

@wf.state
class State(BaseModel):
    query: str
    answer: str = ""

@wf.step
async def think(state: State) -> State:
    return state.model_copy(update={"answer": "..."})

result = wf.run_sync(State(query="Hello"))
print(result.events)  # full per-step timeline
```

## Documentation pages

- Quickstart: https://jamjet.dev/quickstart
- Python SDK guide: https://jamjet.dev/python-sdk
- YAML workflows: https://jamjet.dev/yaml-workflows
- Eval harness: https://jamjet.dev/eval
- MCP integration: https://jamjet.dev/mcp
- A2A protocol: https://jamjet.dev/a2a
- Observability: https://jamjet.dev/observability
- CLI reference: https://jamjet.dev/cli
- Deployment: https://jamjet.dev/deployment
- Compare vs LangGraph/CrewAI/AutoGen: https://jamjet.dev/compare
- Benchmarks: https://jamjet.dev/benchmarks

## Migration guides

- From LangGraph: https://jamjet.dev/migrate/from-langgraph
- From CrewAI: https://jamjet.dev/migrate/from-crewai
- From raw OpenAI SDK: https://jamjet.dev/migrate/from-openai-direct

## Blog

- Why I built JamJet in Rust: https://jamjet.dev/blog/why-rust-for-ai-workflows
- Migrating from LangGraph: https://jamjet.dev/blog/langgraph-migration-real-example
- Self-evaluating agent in 50 lines: https://jamjet.dev/blog/self-evaluating-agent-50-lines
- Testing AI agents like software: https://jamjet.dev/blog/testing-ai-agents-like-software

## Repositories

- Main (Rust + Python SDK): https://github.com/jamjet-labs/jamjet
- Examples: https://github.com/jamjet-labs/examples
- Benchmarks + migration guides: https://github.com/jamjet-labs/jamjet-benchmarks
- Website: https://github.com/jamjet-labs/jamjet.dev

## Contact

- Website: https://jamjet.dev
- GitHub: https://github.com/jamjet-labs/jamjet
- X / Twitter: https://x.com/jamjetdev
- GitHub Discussions: https://github.com/jamjet-labs/jamjet/discussions