Manage multiple, isolated AI coding sessions across VS Code and the terminal. Claude Code, Codex CLI, Antigravity, Gemini CLI, Cortex Code, OpenCode, and Pi out of the box. No context contamination. No broken git states.
Capabilities
Built for developers who run multiple AI coding agents without the chaos of conflicting file changes.
Each session runs in its own Git Worktree. File changes in one lane never leak into another. Switch contexts without stash/pop headaches.
Closing VS Code? No problem. Lanes remembers your active sessions and resumes them exactly where you left off using --resume.
See at a glance which agents are waiting, working, or have errors. Auto-configured hooks keep you informed.
Review all changes with a built-in diff viewer. Compare against any branch and add code review comments before merging.
Create sessions with a form: choose your agent, set a name, starting prompt, attach files, select source branch and permission mode.
Done with a session? Delete the worktree and terminal with one click. Your git branch stays intact and ready to merge.
And more
Platforms
Use Lanes wherever you write code — in your IDE or from the terminal.
Full-featured extension with sidebar, session form, diff viewer, file attachments, and more.
Install from Marketplace →Manage sessions directly from the terminal. No IDE required — perfect for headless and remote workflows.
Build from source →Workflow
Open the Lanes sidebar, choose your agent (Claude Code or Codex CLI), fill in a session name like fix-login-bug, your starting prompt, and optionally attach files.
Lanes creates a hidden Git worktree (.worktrees/fix-login-bug), spawns a dedicated terminal, and launches Claude with your prompt.
Create as many sessions as you need. Each agent works on its own isolated copy of the codebase. No conflicts, no context contamination.
Use the built-in Git Changes viewer to review all modifications. Delete the session when satisfied — worktree removed, branch preserved, ready to merge.
Automation
MCP-based workflows that guide Claude through structured phases: Plan → Implement → Test → Review. Built on Anthropic's research for long-running agent reliability.
name: Feature Development Workflow
description: Structured workflow with plan, implement, test, review
agents:
coder:
description: Responsible for implementing features
test-engineer:
description: Responsible for writing and executing tests
loops:
implement:
- id: programming
agent: coder
instructions: Implement feature...
- id: testing
agent: test-engineer
instructions: Write tests...
steps:
- id: plan
type: action
instructions: Analyze goal and break into features...
- id: implement
type: loop
- id: review
type: action
instructions: Review all changes...Plan → implement → test → review with clear checkpoints
Dedicated agents for coding, testing, and reviewing
MCP integration tracks state across context windows
Setup
Choose your platform and install at least one coding agent.
Lanes is free. Use your own Claude Code subscription (Pro, Team, or Enterprise).
Use your own OpenAI API key to run Codex CLI sessions alongside Claude Code.