Documentation

AgentLane is a Python runtime for agents that send messages, receive events, and keep state across tasks. Use these guides to build an agent app, extend the harness, or run agents across workers.

Start here

To install AgentLane and run your first example, follow the project quickstart.

Choose a guide for the part of your app that you need to build:

• Local agent loops: Start with Default agents.
• Message-driven services: Start with Engine and execution.
• Distributed agents: Start with Distributed runtime usage.
• TypeScript apps: Start with the Process bridge.

For runnable applications, see the example index.

Architecture

The application-facing layers have the following responsibilities:

• Runtime and messaging: Agent addresses, message routing, delivery, and execution.
• Transport: Payload serialization across process boundaries.
• Models: Prompts, tools, schemas, and model calls.
• Harness: Agent loops, handoffs, and resumable runs.
• Tracing: Execution records across all layers.

Use a local runtime for one process. Use a distributed runtime when you need cross-worker routing or worker placement. Both use the same public messaging model.

Contents

Use the following references to explore each part of AgentLane.

Runtime

Create local runtimes and distribute work across hosts:

• Engine and execution
• Distributed usage
• Distributed architecture

Messaging

Route messages and track delivery:

• Routing and delivery

Transport

Serialize payloads for transport:

• Serialization

Models

Configure model calls and prompts:

• Overview
• Prompt templating

Harness

Build agent loops and extend their behavior:

• Architecture
• Tasks
• Agents
• Default agents
• Shims
• Compaction
• Tools
  • Design
  • Permissions
  • MCP
  • agent
  • bash
  • find
  • grep
  • patch
  • read
  • write
  • write_plan
• Skills
• Markdown agent definitions
• Runner
• Run event serialization
• Distributed agents
• File I/O interfaces

Sandboxes

Run environment tools in a sandbox managed by your app:

• Introduction
• SPACE

Process bridge

Connect a TypeScript app to a Python backend:

• Overview
• Runtime configuration
• Protocol and lifecycle
• Development

Tracing

Inspect execution with tracing:

• Overview

Project

Read project history and contribution guidance:

• Changelog
• Code style
• Maintain documentation