AI Assistant Integration

RoomKit publishes its documentation in forms a coding assistant (Claude Code, Cursor, Copilot…) can read, so the code it writes uses RoomKit's actual API rather than a guess from its training data.

The files

File	What it holds	Where
llms.txt	Index of the documentation, one line per page, in the llms.txt format	www.roomkit.live/docs/llms.txt, and inside the installed package
llms-full.txt	The topic pages (architecture, channels, hooks, voice, orchestration, API) in one file	www.roomkit.live/docs/llms-full.txt, and inside the installed package
AGENTS.md	Conventions for working on RoomKit itself: commands, project layout, patterns, code style, what to ask first	repository root, and inside the installed package

Give the assistant llms.txt or llms-full.txt to build an application with RoomKit; AGENTS.md is for contributing to RoomKit.

From Python

The package ships the copies that match the installed version:

from roomkit import get_agents_md, get_ai_context, get_llms_full_txt, get_llms_txt

index = get_llms_txt()          # llms.txt
full = get_llms_full_txt()      # llms-full.txt
guidelines = get_agents_md()    # AGENTS.md
context = get_ai_context()      # AGENTS.md and llms.txt in one string

Agent Skills

roomkit-skills holds skills that walk an assistant through common tasks: setting up a project, creating a room, an AI channel, a voice agent, an SMS or WhatsApp channel, hooks, orchestration, PostgreSQL and telemetry. For Claude Code:

claude plugin marketplace add https://github.com/roomkit-live/roomkit-skills
claude plugin install roomkit-dev

MCP

RoomKit's AI channels can call tools from MCP servers; see MCP Integration.