Unify email, chat, tasks, and CRM into a single context for teams and AI
Experience a workflow where you read a customer email, create a task from it, and have an agent process that task into a document, all within one screen. Designed to prevent context loss when switching between multiple tools, this system connects all team communication and records into a single graph. As a result, both team members and AI agents can work while sharing the same information.
Reducing confusion caused by tool fragmentation
Previously, teams had to use multiple tools separately, such as Slack, Linear, Notion, and HubSpot, but this repository integrates them into a single system. Each function operates independently, yet all items are stored as a bidirectional graph where they reference one another. This allows you to track the process from an email mention to a task, and from a task to a pull request, at a glance. As the team grows, it helps reduce confusion caused by fragmentation between tools and enables the entire company to operate as a single computable system.
Functional areas composed of blocks
The workspace is composed of modular blocks, with each block optimized for specific tasks.
- Unified email Process emails from multiple accounts in a single inbox and navigate quickly using keyboard shortcuts. You can create tasks directly from emails or share them to channels.
- Connected chat Channels and direct messages are bidirectionally linked to tasks, documents, and CRM records. You can also bring external contacts or WhatsApp conversations into this system.
- Context-based tasks Create tasks directly from chats or emails, which are automatically linked to the generated context. When you assign a task to an agent, it performs the work based on relevant information and updates the status.
- Collaborative documents and databases In Markdown-based documents, you can link all workspace items using @mentions. Database tables are connected to people, tasks, and files, updating in real time.
Agents with team-wide memory
Agents operate based on the team's entire email, document, task, and call content, rather than individual chat records. This enables agents to understand the background of requests and perform more accurate work. Agents can write and execute code via the SDK to create custom workflows. Additionally, it supports the MCP protocol, allowing external coding agents like Claude Code or Codex to be connected to and used within this workspace.
Considerations before adoption
This project is released under the GNU Affero General Public License v3.0 (AGPLv3) and is fully open source. While self-hosting is possible, a separate inquiry is required for deployment under a commercial license. In terms of security, it has received SOC 2 Type II certification and explicitly states that data residency with model providers is zero. Before adoption, you should review the switching costs associated with your team's existing tools and the legal implications of the AGPLv3 license. If you wish to run it locally or contribute, refer to the 'Running locally' section in the documentation.
