Pre-built IDE and coding-agent integrations

[haasonsaas]

Problem

Room exposes a generic hook framework (roomctl hook pre-tool) and an MCP sidecar, but there are no turnkey integrations for specific coding agents. Developers using Cursor, Claude Code, GitHub Copilot, Windsurf, or similar tools must manually wire hooks and configure MCP endpoints.

Proposal

Provide first-class, documented integrations for the most common coding agents, each with:

• Install/setup scripts or config templates
• Hook definitions that map agent lifecycle events to Room's pre-tool / post-tool hooks
• MCP configuration snippets for the room-mcp sidecar
• A short integration guide per agent

Priority order (based on adoption):

1. Cursor
2. Claude Code
3. GitHub Copilot
4. Windsurf / others

Why

The generic hook framework is powerful but adoption depends on making the most common agents work with near-zero manual configuration.

[Keesan12]

This is the right direction, especially the move from vague "agent support" to explicit lifecycle hooks.

One thing I would make first-class in the integration contract is the difference between tool invocation and run governance. A lot of coding-agent integrations stop at "here is how to call tools," but the operational problems usually show up one layer higher: what paused the run, what was auto-approved, what was retried, what evidence was produced, and how a host resumes safely after interruption.

If Room becomes the place where those hooks normalize across agents, the higher-value surface may be a small common run receipt schema in addition to the setup templates. That makes post-run eval, replay, and policy checks much easier than scraping per-agent logs later.