Piclaw
pi coding agent in a technicolor web trenchcoat
Last verified:
What is Piclaw?
PiClaw is a self-hosted AI workspace that packages the Pi Coding Agent into a Docker container with a streaming web UI. It provides a unified environment combining chat, code editor, terminal, file browser, document viewers, draw.io, kanban boards, VNC client, and MCP access — all accessible through a single docker run command. The tool supports multi-provider LLMs including Anthropic, OpenAI, Azure, Gemini, Ollama, and any OpenAI-compatible endpoint.
Key features include streaming chat with Markdown/KaTeX/Mermaid support, persistent state backed by SQLite (covering messages, media, tasks, token usage, and an encrypted keychain), workspace tooling with CodeMirror 6 editor supporting 12 languages, Office/PDF/CSV/image/video viewers, and infrastructure tools like SSH, Proxmox, and Portainer profiles. The surface area is effectively infinite thanks to a growing catalog of community add-ons covering Proxmox, Portainer, SSH, and more.
PiClaw is designed for developers, DevOps engineers, and AI enthusiasts who want one stateful agent workspace they can run locally or in a container without stitching together half a dozen separate services. It supports agent-first workflows including steering, queued follow-ups via /btw, side prompts, autoresearch loops, scheduled tasks, and visual artifact generation. Authentication uses WebAuthn passkeys with TOTP as fallback, and optional WhatsApp integration is available.
Piclaw pricing
Pricing model: Freemium
PiClaw is free and open-source under the MIT license. It is self-hosted with no paid tiers — users run it locally via Docker (ghcr.io/rcarmo/piclaw:latest) or install without Docker using bun add -g github:rcarmo/piclaw. The only costs are your own infrastructure (server/VPS/hardware) and LLM API costs from your chosen provider (Anthropic, OpenAI, Azure, Gemini, Ollama, or custom OpenAI-compatible endpoints). No subscription fees, credit systems, or usage limits from PiClaw itself.
Piclaw pros
- Single Docker container deployment with one docker run command
- Multi-provider LLM support (Anthropic, OpenAI, Azure, Gemini, Ollama)
- Streaming chat with token-by-token SSE responses
- Built-in Ghostty-based web terminal (detachable popouts)
- CodeMirror 6 code editor with 12-language syntax highlighting
- SQLite-backed persistent state for messages, media, tasks
- Encrypted keychain for secret storage
- Office/PDF/CSV/image/video viewers in one interface
- draw.io integration for diagrams
- Built-in kanban boards for project tracking
- VNC client for remote machine access without leaving browser
- SSH, Proxmox, and Portainer profile support for infrastructure
- CDP browser automation capabilities
- MCP (Model Context Protocol) access
- Community add-on catalog for extended functionality
- WebAuthn passkey authentication with TOTP fallback
- Optional WhatsApp integration
- Autoresearch loops and scheduled task support
- Mobile-friendly responsive web UI
- Multi-arch Docker images (amd64 and arm64)
Piclaw cons
- Single-user only (no multi-tenant support)
- Windows support is secondary/not officially supported
- Requires Docker or Bun runtime (not truly standalone)
- Must manually preserve /workspace/.piclaw/store/messages.db
- TOTP secret must be configured manually for auth
- No built-in credit/usage billing (self-hosted only)
- Keychain master key must be set via environment variable
- Experimental M365 extension bundle (not production-ready)
- No official cloud-hosted version available
Frequently asked questions about Piclaw
How do I start PiClaw with Docker?
Run mkdir -p ./home ./workspace then docker run -d --init --name piclaw --restart unless-stopped -p 8080:8080 -e PICLAW_WEB_PORT=8080 -v "$(pwd)/home:/config" -v "$(pwd)/workspace:/workspace" ghcr.io/rcarmo/piclaw:latest. Open http://localhost:8080 and type /login to configure your LLM provider.
What LLM providers does PiClaw support?
PiClaw supports Anthropic, OpenAI, Azure, Gemini, Ollama, and any OpenAI-compatible endpoint. You configure provider credentials via /login in the chat — API keys are not set as environment variables. Custom OpenAI-compatible endpoints work when not using built-in hosted providers.
How does authentication work in PiClaw?
PiClaw uses WebAuthn passkeys for authentication with TOTP as fallback. Set PICLAW_WEB_PASSKEY_MODE to totp-fallback (default), passkey-only, or totp-only. Enable TOTP by setting PICLAW_WEB_TOTP_SECRET as a Base32 secret or initialize with /totp command in chat.
Where does PiClaw store persistent data?
Home state (.pi/, .gitconfig, .bashrc) mounts to /config on the host. Workspace (projects, notes, state) mounts to /workspace. Chat history, media, and task state live in /workspace/.piclaw/store/messages.db — never delete this SQLite file or you lose session data.
What code editor does PiClaw use?
PiClaw uses CodeMirror 6 with search/replace, dirty-state tracking, and syntax highlighting for 12 languages. The editor supports lazy local bundles and is accessible directly in the web UI without separate apps.
Can I run PiClaw without Docker?
Yes, experimentally. Install with bun add -g github:rcarmo/piclaw on Linux/macOS/Windows. See docs/install-from-repo.md for the Docker-free install path that gives you the packaged PiClaw CLI and web assets without building from source manually.
What automation tools are built into PiClaw?
PiClaw includes /image, /flux, image_process for image handling, cdp_browser for browser automation, mcp for Model Context Protocol, experimental m365 for Microsoft 365, and Windows-only win_* tools. There's also a skills system extending into Playwright automation, web search, charting, and task scheduling.
How do I enable the web terminal?
Set PICLAW_WEB_TERMINAL_ENABLED=1 (default on Linux/macOS, 0 on Windows). The Ghostty-based web terminal runs as a dock or tab with detachable popouts. It requires authentication and is session-scoped with SSH-backed remote tools.
What are autoresearch loops?
Autoresearch loops are autonomous experiment loops that let PiClaw conduct iterative research. The feature comes from pi-autoresearch by Tobi Lutke and David Cortés, now carried by the autoresearch add-on in rcarmo/piclaw-addons. It enables the agent to search, summarize, and iterate on research tasks automatically.
Is PiClaw affiliated with pi.dev?
No, PiClaw is not directly affiliated with pi.dev. It is a derivative work that leverages Pi's core functionality and builds additional runtime, tooling, and UI layers around it. The project credits pi.dev for the flexible core but is maintained independently by rcarmo.