Documentation

JamJet

Build durable AI agents with checkpoint replay, traces, runtime-enforced limits, and protocol-native interoperability.

Get started Browse docs

Start in 60 seconds See crash recovery Compare frameworks Migrate from LangGraph

$pip install jamjet

Get started

Quickstart

First agent in 5 minutes

→

Concepts

Architecture and mental model

→

CLI Reference

jamjet init, dev, run, validate

→

SDKs

Python SDK

@task, Agent, Workflow

→

Java SDK

JDK 21, virtual threads, records

→

YAML Workflows

Declarative workflow definitions

→

Protocols

MCP Guide

Connect to any tool server

→

A2A Guide

Agent-to-agent communication

→

Research & Eval

Research Guide

Experiments, grids, publication export

→

Eval Harness

Scorers, judges, CI gates

→

Operations

Observability

OTel traces, cost attribution

→

Deployment

Docker, Kubernetes, Fly.io

→

Enterprise

Policy, PII, multi-tenant, mTLS

→