pikiloom CLI
pikiloom runs Claude Code, Codex and Gemini from one web dashboard and chat apps so you can steer agents from a phone.
- 293 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The agent runs commands on your machine with access to local files and tools and can be steered from a chat app.
- Stores provider keys and bot tokens locally, and someone else's access to the bot opens control of the agent.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/pikiloomDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent runs commands on your machine with access to local files and tools and can be steered from a chat app.
- Stores provider keys and bot tokens locally, and someone else's access to the bot opens control of the agent.
This is third-party code. Review the repository files before installing.
What it does
pikiloom is a CLI orchestrator started with `npx pikiloom@latest` that opens a web dashboard at `http://localhost:3939`. From the dashboard you run parallel Claude Code, Codex, Gemini and Hermes sessions, connect models through providers and profiles, and install skills and MCP servers. The same sessions are reachable from Telegram, Feishu, WeChat, Slack, Discord, DingTalk and WeCom. Installed skills are invoked in any connected terminal with `/sk_<name>`. The agent works on your machine with access to local files and tools rather than in a vendor sandbox.
Who it is for. Developers running several agents at once who want to control them from a chat app while away from the computer.
Good fit when
- Claude Code writes code in one pane and Codex runs tests in another, with both sessions on one screen.
- A long agent task must be checked and steered from Telegram without opening a laptop.
- A session started in Claude is handed to Gemini for review without losing context.
- A team in Slack or Feishu steers one agent together in a shared group.
Not a fit when
- One agent in a terminal is enough and chat apps are not needed.
- The agent must not get access to the whole local disk and tools.
- You want an isolated vendor cloud sandbox rather than running on your own machine.
Example request
Start the pikiloom dashboard, connect a Telegram bot, give Claude Code a task to write tests and ask Gemini to review.Limitations
Node.js 20 or newer is required. The Claude Code, Codex, Gemini and Hermes agents are installed separately, and pikiloom manages those already present. Provider keys are stored in `~/.pikiloom/setting.json`. Chat apps need their own bot tokens, for example `TELEGRAM_BOT_TOKEN` for a Docker run on a server. The macOS computer-use mode through the `peekaboo` MCP is enabled manually and asks for Accessibility and screen recording permissions.
How to disable. Stop the `pikiloom` process and disable the bots in the chat apps. Remove data and provider keys in the `~/.pikiloom` directory.
FAQ
How do I start pikiloom?
Run `npx pikiloom@latest` and the web dashboard opens on port 3939. The `--setup` option starts an interactive wizard in the terminal and `--doctor` only checks the environment.
Which agents are supported?
Drivers for Claude Code, Codex, Gemini and Hermes are built in. The registry accepts any CLI or ACP agent through the shared `AgentDriver` contract.
Can it run on a server?
Yes. The README has a Docker run with a bot token such as `TELEGRAM_BOT_TOKEN`. Take Docker parameters from the Quick Start section.
Where do project skills live?
Project skills are loaded from `.pikiloom/skills/*/SKILL.md`, and the legacy `.claude/commands/*.md` format is supported too.
Where are provider keys stored?
Credentials are isolated in `~/.pikiloom/setting.json`. Profiles that bind a model to an agent take them from there.
Discussion
No questions or reviews yet.
Related
CLI toolsA compact OpenClaw alternative: a personal assistant on Claude Agent SDK with agents in isolated containers and messengers added via skills
A phone as a remote for Claude Code, Codex and OpenCode: watch the agent work and approve actions while it stays on your PC.
One browser window for Claude Code, Codex, Gemini CLI and OpenCode: session status, history and requests between agents.
A CLI that lets agents in different terminals message, watch and spawn each other
An AI-agent workspace with parallel worktrees, cross-provider orchestration, and bots across chat channels.
Magpie
magpie
Magpie — an app and local gateway that connect any model to Claude Code, Codex, Gemini CLI and OpenCode and switch them from the menu bar
README in short
The orchestration core joins four layers: entry points (chat apps, web dashboard, API and CLI), agents, model routing and a toolset of skills, MCP and CLI.