Caas — AI skill for Claude Code
CAAS (Chat–Agent Auto Session): talk only to your everyday Claude.ai conversation; running Claude Code sessions do the work and wake the same conversation with the result.
How to install Caas
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open Rurimpa/caas and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Caas does
CAAS (Chat–Agent Auto Session): talk only to your everyday Claude.ai conversation; running Claude Code sessions do the work and wake the same conversation with the result. Includes a tool-limiting MCP gate.
Alternatives in AI
- Ralph For Claude Code — by Frank Bria - An autonomous AI development framework that enables Claude Code to work iteratively on project 8k ★
- Open Multi Agent — TypeScript multi-agent orchestration engine — one runTeam() call from goal to result 5.8k ★
- Agentchattr — Free, local chat where AI coding agents can tag each other, talk, and coordinate with you 1.5k ★
README
CAAS — Chat–Agent Auto Session
⚠️ **Use CAAS only with the gate (`code/gate/caas_gate.js`) in place.** Without it, anyone who can talk to the chat can run any command on the PC. See [Safety](#safety).
**CAAS lets a person talk only to their everyday Claude.ai conversation, while Claude Code sessions that are already running on the same PC do the work, and the results come back to the same conversation without any further human input.**
Technical note (DOI): [10.5281/zenodo.23005933](https://doi.org/10.5281/zenodo.23005933) · Author: がけっぷちのふくしゃちょう / The Cliff-Edge Fukushacho
What is new
Parts of this already exist in public: sending from a chat to a running Claude Code session, and waking a bound session with a scheduled trigger. But every public bridge we found needs a human action on the chat side before the loop closes (typing `get`, asking "how is it going?", or waiting inside a single human-initiated turn).
CAAS closes the loop without a human:
| CAAS | Dispatch (official) | cc-tap / mcp-relay / herald | |
|---|---|---|---|
| Human talks to | their everyday Claude.ai conversation | a dedicated Dispatch conversation | everyday conversation |
| Work goes to | Claude Code sessions already running | newly spawned sessions | running sessions (herald: new) |
| Result returns without a human | yes | inside Dispatch only | no |
Why "everyday" and "already running" matter: the everyday conversation is where the person thinks (history, memory, projects); the running sessions carry accumulated context. Consolidating the human's window means driving *those*, not a separate front desk with temporary workers.
How it works
[everyday chat] --(1) MCP message--> [gate] --> [running Claude Code session]
^ |
| (2) write result to a file
| (3) run the bound trigger
+------(4) the bound
Related Skills
Icor For Life Chat
Talk to your AI team where the knowledge lives. A Claude chat inside the vault: it sees the note you have open
Agents Can Communicate
Let your AI coding sessions talk to each other. Coordinate parallel work, exchange reviews, and hand off unfin
Context Summary
A Claude.ai Personal Skill that generates a structured, seven-block session handoff so you can continue work i
Team Build
Experimental agent-teams build where this session acts as team lead (escalated model) coordinating everyday-mo
Venice Py
Unofficial async-first Python SDK for the Venice.ai API — chat, image, video, music, audio, embeddings, TEE at
Agents In A Chest.Nvim
Neovim plugin to run LLM coding agents in isolated Git work trees with auto‑commit, tabbed sessions, and safe
Related Agents
GPT Chat
Conducts a conversation with a GPT model (Codex) as a separate conversational partner—the thread persists betw
Wip Manager
Manage work-in-progress for multi-session tasks. Auto-invoked when tasks span sessions.
Conversation Historian
Use when analyzing Claude Code conversation history, finding past sessions by topic, tracing how implementatio