LanesAgents
Coding agents that work with Lanes
In shortLanes works with 19 coding agents. Twelve get its hooks or a plugin, installed with lanes init --agents, which coordinate every edit: Claude Code, Codex, Gemini CLI, Cursor, GitHub Copilot, Kiro, Antigravity, Factory Droid, Qwen Code, opencode, Kilo Code and Amp. Seven more get the team tools through the MCP server: Augment, Junie, JetBrains AI Assistant, Continue, Zed, Goose and Warp. Teams can mix agents freely.
Agents with hooks or plugins
lanes init --agents <names> installs the integration for each agent you name, comma-separated, and commits it to the repository. For example: lanes init --agents claude,codex,cursor.
| Agent | Integration | A teammate on the same file | Team brief at the prompt | Status |
|---|---|---|---|---|
| Claude Code | Hooks in .claude/settings.json | Queued before the edit | Yes | Run live |
| OpenAI Codex | Hooks in .codex/hooks.json | Queued before the edit | Yes | Built from docs |
| Gemini CLI | Hooks in .gemini/settings.json | Queued before the edit | Yes | Built from docs |
| Cursor | Hooks in .cursor/hooks.json | Published after the edit | Through MCP | Built from docs |
| GitHub Copilot | Hooks in .github/hooks/lanes.json | Published after the edit | Through MCP | Built from docs |
| Kiro | Hooks in .kiro/hooks/lanes.json | Published after the edit | Through MCP | Built from docs |
| Google Antigravity | Hooks in .agents/hooks.json | Queued before the edit | Through MCP | Built from docs |
| Factory Droid | Hooks in .factory/hooks.json | Queued before the edit | Yes | Built from docs |
| Qwen Code | Hooks in .qwen/settings.json | Queued before the edit | Yes | Built from docs |
| opencode | Plugin in .opencode/plugins/lanes.js | Queued before the edit | Through MCP | Built from docs |
| Kilo Code | Plugin in .kilo/plugin/lanes.js | Queued before the edit | Through MCP | Built from docs |
| Amp | Plugin in .amp/plugins/lanes.ts | Queued before the edit | Yes | Built from docs |
"Queued before the edit" means the agent's edit waits about a second for a teammate's edit of the same file, then goes ahead on the current version. "Published after the edit" means the agent's edit is committed and shared right away, and teammates' agents are queued and warned about it; the agent's own hooks cannot wait before an edit.
Agents through MCP only
These agents have no hook Lanes can use, so they get the team tools through lanes mcp. Their edits are not queued or published by Lanes, but the agent can read the brief, check a file before editing it, post to the team chat and send requests.
| Agent | Where the MCP config goes |
|---|---|
| Augment (Auggie CLI and editors) | ~/.augment/settings.json |
| Junie | ~/.junie/mcp/mcp.json |
| JetBrains AI Assistant | Settings, Tools, AI Assistant, MCP, Add, As JSON |
| Continue | ~/.continue/config.yaml |
| Zed | settings.json, under context_servers |
| Goose | goose configure, Add Extension, Command-line Extension |
| Warp | ~/.warp/.mcp.json |
The exact snippet for each is in the MCP reference.
Mixing agents on one team
A team does not need to agree on one agent. Lanes identifies people, not tools: the brief says "Ana's agent is renaming getUser" whether Ana runs Claude Code or Cursor. Each person installs the integration for the agents they use, and lanes init --agents can list several at once.
Questions
Can teammates use different agents?
Yes. Every agent reports to the same Lanes server, so a Claude Code user and a Cursor user see each other's work in the brief, the heads-ups and the views.
Which agent works best with Lanes?
Claude Code. It is the agent Lanes was run against in every benchmark, and it gets every feature, including the read heads-up and the end-of-turn check.
What does "built from docs" mean?
The integration follows the agent's published hook or plugin documentation and is covered by Lanes' automated tests, but it has not yet been run against the real agent. Reports are welcome on GitHub.