English · 简体中文
Download · Self-host · Development · Contributing
CogineBot is a self-hosted application for persistent AI Bots, forked from Rakazo. It supports ongoing conversations, group chats, Bot-to-Bot messaging, files, and a computer workspace. You can choose your models and use the web or desktop client.
We were among Rakazo's first users, and we're grateful to its author for making the project open source.
Personal agents are still at an early stage, and there are many ideas about what they should be and how they should work. We have our own ideas about design and product direction, and we'd like to explore them through this fork.
CogineBot is maintained independently by cogine-ai. So far, we have reworked the chat timeline, sidebar, group interactions, avatars, files, and computer takeover flow. The adaptation notes describe these changes. We follow upstream and bring over changes that fit.
- Persistent Bots with conversation history, memory, and routines.
- Portable Bot packages with reviewed JSON imports and exports.
- Group chats with multiple Bots, @ mentions, and Bot-to-Bot messages.
- File attachments, previews, version history, and downloads.
- Browser, terminal, and file access through a Bot's computer. Watch first, select Take control to interact, and I'm done to return control.
- Configurable models, application connections, and computer providers.
Optional services need to be configured on your instance. See the self-hosting guide.
Download installers and checksums from GitHub Releases.
| Platform | Package | Signing |
|---|---|---|
| macOS, Apple silicon and Intel | Universal .dmg / .zip |
Signed and notarized |
| Windows x64 | .exe installer |
Currently unsigned; Windows may show an unknown-publisher warning |
| Linux x64 | .AppImage |
Checksums and build provenance included |
Choose This computer to run a local instance with Docker Engine 26+ and Compose, or Existing instance to connect to a server you already run. See desktop release details.
The repository also includes the Expo mobile client. CogineBot does not yet have its own public mobile release or hosted service.
Follow the Docker Compose installation guide. It covers configuration, optional providers, backups, and updates. The web and desktop clients connect to the same instance.
Use Node.js 24 and pnpm 9.15.0. The supported Node range is in package.json.
git clone https://github.com/cogine-ai/CogineBot.git
cd CogineBot
node scripts/check-repository.mjsThis checks repository documentation and workflow metadata. For dependencies, tests, and running the application, see Development.
Current checks are in GitHub Actions. The preparation baseline records the initial checks; agent verification describes the test harnesses and what they cover.
- Connectors: improve application connections and connected-account management.
- iOS client: test connections to self-hosted servers and everyday use through TestFlight, then prepare an App Store release.
Issues and pull requests are welcome. For substantial changes, please open an issue first to discuss the scope.
- Bug reports: include the version and steps to reproduce.
- Feature requests: describe the use case. Requests for specific application connections are welcome here too.
- Setup help: desktop, server, or source installation questions.
See CONTRIBUTING.md for development and review guidelines. Report security vulnerabilities through the private reporting channel.
UPSTREAM.md records the source revisions and integration policy. Internal @rakazo/* package names and protocol/storage identifiers are retained for compatibility. The original README.upstream.md is preserved verbatim. See Automation for this repository's checks and release workflows.
CogineBot preserves Rakazo's history from 794a76da6eb0a73532a89cf57d2f202ef9b6b6c8, its Apache-2.0 license, and existing copyright and attribution notices. Additional terms for identified copied components are in THIRD_PARTY_NOTICES.md; this is not a complete dependency or asset inventory.
Modified on 2026-10-16 for CogineBot's project introduction, fork rationale, distribution, and contribution information. Original upstream documentation is preserved separately.