Crow

Crow

AI-powered project management with persistent memory, encrypted P2P sharing, and 20+ integrations, enabling your AI assistant to manage projects, share memories, and collaborate securely across platforms.

Category
Visit Server

README

Crow

Crow is a modular, agentic framework for personalized assistance, project management, research, education, home entertainment, and more. It works with the AI tools and services you already use, and it runs on almost any device, from a Raspberry Pi to a free cloud server to your desktop. Crow's multi-instance design pulls all of your devices into one private, user-friendly interface where you build and run your own agents. It brings you the full power of modern AI while letting you reclaim your privacy and keep your data out of big tech's hands.

Built on the open Model Context Protocol standard. Published by Maestro Press | Product Page | Docs

Your AI, your devices, your data

Most AI products are a window into someone else's server. Crow inverts that. You run the server, on hardware you own, and your memories, projects, files, and agents live in a local database that never has to leave.

Privacy is honest here, not absolute marketing. Crow stores your data on infrastructure you control. Pair it with a local model (Ollama or any OpenAI-compatible endpoint on your own machine) and your data never leaves your network. Connect a cloud assistant (Claude, ChatGPT, Gemini, and others) and only what you choose to send that provider goes out, on your terms. Either way, the system of record is yours.

Build and run your own agents

The Bot Builder is the spine of Crow's agentic side. An agent (a "bot") is a persona plus the skills, tools, gateways, and permissions you give it, all configured from the dashboard with no config files to hand-edit.

  • Compose an agent: Give it a name and personality, pick its model, attach skills, and select exactly which tools it can use (Crow's own memory/projects/blog/storage tools, plus any installed extension's tools).
  • Put it on a channel (gateways): The same agent can answer email (Gmail), chat on Discord, or run hands-free on Meta Ray-Ban glasses as a fast voice assistant. One agent, the channels you choose.
  • Scoped and permissioned by default: Each agent only sees the tools you grant it. A permission_policy governs what it may do without asking: confirm-before-acting, deny outright, or downgrade outbound sends and publishes to drafts. Voice turns enforce the same policy on the underlying action, not just the surface tool name.
  • Opt-in self-authoring: If you turn it on (it is off by default), an agent can draft a new skill for itself into a confined staging area. Nothing takes effect until you review the text and approve it in the dashboard. Agents cannot grant themselves tools or loosen their own permissions: skills are prompt text only.

This is a local-first, secure alternative to cloud bot platforms. Where engines like OpenClaw or Hermes lean on auto-authoring and hosted control, Crow keeps the engine on your hardware and puts an operator-approval gate in front of anything an agent writes for itself.

Bot Builder Guide · Architecture · Meta Glasses

What Crow does

Capability What you get
Persistent memory Your assistant remembers across sessions and platforms. Full-text search, categories, importance scoring, automatic recall.
Projects & research Typed project workflows, multi-format citations (APA, MLA, Chicago), bibliographies, and source verification. Every claim links to a stored source.
Agents (Bot Builder) Build personas, attach tools and skills, and run them over Gmail, Discord, or voice. Scoped, permissioned, operator-approved.
Encrypted P2P sharing Share memories, projects, and messages directly between Crow users. End-to-end encrypted, no central server, no accounts.
Blog & publishing Markdown blog with RSS/Atom feeds, themes, and public URLs. Publish by telling your AI.
File storage S3-compatible storage with quotas and presigned URLs.
20+ integrations GitHub, Slack, Notion, Gmail, Trello, Discord, Google Workspace, and more, proxied through one authenticated gateway.
Self-hosting add-ons Ollama, Nextcloud, Immich, Home Assistant, Jellyfin, and more, installable by asking your AI.
Multi-instance Run Crow on several devices and sync the pieces that should travel with you.

P2P Sharing: A First for AI Platforms

Crow is the first AI platform with built-in encrypted peer-to-peer sharing. No cloud middleman, no accounts to create, just your Crow ID.

  • Share memories and projects: Send a memory or an entire project space to a friend's Crow, encrypted end-to-end.
  • Collaborate on project spaces: Project spaces have members, roles (owner / editor / viewer / guest), and per-member capability overrides. Clone-share delivers a one-shot snapshot today; live one-way subscription is a planned follow-on.
  • Encrypted messaging: Send messages between Crow users via the Nostr protocol with full sender anonymity.
  • Works offline: Shares queue up and deliver when both peers are online. Peer relays handle async delivery.
  • Zero trust: No central server sees your data. Invite codes, safety numbers, and NaCl encryption throughout.

"Clone my project to Alice as a viewer" and that is it. Crow handles the cryptography, discovery, and bundle assembly.

Learn more: Sharing Guide · Architecture

Crow for your field

The same framework points in very specific directions depending on what you do. Each cut below describes capabilities Crow ships today.

  • Education: A private AI workspace for students, teachers, and researchers. Crow remembers your work across sessions, manages sources with real citations and bibliographies, and connects to course tools and public datasets. Built for FERPA-sensitive settings, your data stays on infrastructure you control.
  • Law: A self-hosted agentic assistant for legal work: matter organization, document research, and citation tracking, with privileged material kept on hardware you own. No third-party AI vendor sees the file.
  • Home entertainment: Turn a home server into an agentic hub for your media. Voice-controlled music and video across your devices, even your glasses, drawing on your own library. Your collection, your rules, nothing tracking your habits.
  • Enterprise: A self-hosted agentic platform for teams that handle regulated data. Build internal assistants, connect the tools your team already uses, and keep every byte inside your own network, with operator-approval gates on anything an agent does.

Works With

Claude ChatGPT Gemini Grok Cursor Windsurf Cline Claude Code
Web, Mobile, Desktop Apps/Connectors CLI, AI Studio Remote MCP IDE IDE VS Code CLI

Crow's Nest

Server-rendered web UI with Dark Editorial design. Password-protected, session-based auth. Built-in panels for the Bot Builder, Messages, Blog, Files, Extensions, and Settings. Third-party panels can be installed from ~/.crow/panels/.

Crow's Nest Guide · Architecture

AI Chat Gateway (BYOAI)

Use the Crow's Nest as a chat frontend with your own AI provider: OpenAI, Anthropic, Google, Ollama, or any OpenAI-compatible endpoint. Tool calling routes through Crow's MCP servers, so your AI can reach memories, projects, and files during conversations. No API keys leave your server.

Chat Architecture

Quick Start

Managed Hosting

A pre-configured Crow instance with no setup and no maintenance.

Get managed hosting →

Oracle Cloud Free Tier (Recommended Free)

A permanent free server that never sleeps, with local SQLite and no external database needed.

  1. Create a free Oracle Cloud account
  2. Launch an Always Free VM.Standard.E2.1.Micro instance (Ubuntu 22.04)
  3. Install Crow + Tailscale, create a systemd service
  4. Connect from any AI platform

Full Oracle Cloud guide

Desktop (Claude Desktop)

git clone https://github.com/kh0pper/crow.git && cd crow
npm run setup
npm run desktop-config  # Copy output to Claude Desktop config

Desktop setup guide

Developer (Claude Code)

cd crow
npm run setup
claude  # Loads .mcp.json + CLAUDE.md automatically

Claude Code guide

Raspberry Pi / Self-Hosted (Crow OS)

curl -fsSL https://raw.githubusercontent.com/kh0pper/crow/main/scripts/crow-install.sh | bash
crow status

Installs Crow as a persistent service with the crow CLI for managing bundles and updates. Supports Raspberry Pi, Debian, and Ubuntu.

Full setup guide

Crow OS & Self-Hosting

Crow OS turns a Raspberry Pi or any Debian machine into a personal AI server. The crow CLI manages the platform and installable add-on bundles:

  • crow status: Platform health, identity, and resource usage
  • crow bundle install <id>: Install add-ons like Ollama, Nextcloud, or Immich
  • crow bundle start/stop/remove: Lifecycle management for bundle containers

Self-hosting add-ons include local AI (Ollama), file sync (Nextcloud), photo management (Immich), smart home (Home Assistant), and knowledge management (Obsidian).

Crow OS Installer · Add-on Registry

Developer Program

Crow is open to contributions. Build integrations, skills, tools, and deployment bundles for the ecosystem.

  • MCP Integrations: Connect new services (Linear, Jira, Todoist, etc.)
  • Skills: Write behavioral prompts that teach the AI new workflows (no code required)
  • Core Tools: Add MCP tools to crow-memory, crow-projects, crow-sharing, crow-storage, or crow-blog
  • Self-Hosted Bundles: Create Docker Compose configs for specific use cases

Developer Docs · Community Directory · CONTRIBUTING.md

Documentation

Full documentation at maestro.press/software/crow

  • Bot Builder: Build agents with personas, tools, skills, gateways, and permissions
  • Meta Glasses: Run an agent hands-free on Ray-Ban Meta glasses
  • Managed Hosting: Pre-configured Crow instance, no setup required
  • Platform Guides: Setup for Claude, ChatGPT, Gemini, Grok, Cursor, Windsurf, Cline
  • Integrations: All 20+ services with API key setup instructions
  • Sharing & Social: P2P encrypted sharing, messaging, and collaboration
  • Storage: S3-compatible file storage with quotas and presigned URLs
  • Blog: AI-driven publishing with themes, RSS, and public sharing
  • Crow's Nest: Web UI with panels for the Bot Builder, messages, files, blog, and extensions
  • Architecture: System design, server APIs, gateway details
  • Skills: Behavioral prompts for AI workflows
  • Security: API key safety, deployment security, and what to do if a key leaks
  • Troubleshooting

License

MIT

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured