rowboatlabs/rowboat

Open-source AI coworker, with memory

What it solves

Rowboat addresses the problem of AI assistants that "start cold" by reconstructing context on demand. Instead of relying on simple retrieval, it creates a compounding, long-lived memory of your work that is stored locally and remains inspectable and editable by the user.

How it works

Rowboat indexes data from emails, meetings, Slack, and assistant conversations into a backlinked knowledge graph. It stores all information as plain Markdown files on the user's machine. This local-first approach allows the AI to maintain explicit relationships between pieces of information over time. Users can connect their own preferred models (local via Ollama/LM Studio or hosted via API) and extend the AI's capabilities using the Model Context Protocol (MCP) to connect to external tools like CRMs, databases, and search engines.

Who it’s for

Knowledge workers and developers who need an AI coworker that remembers their work history, manages communications, and can execute tasks across different applications.

Highlights

  • Integrated Work Surfaces: Includes a built-in email client, browser, meeting note-taker, and code mode.
  • Bespoke Memory: Uses a living knowledge graph rather than standard RAG to maintain long-term context.
  • Local-First: All data is stored as plain Markdown, preventing vendor lock-in.
  • Agentic Capabilities: Supports background agents that trigger on events or schedules to write code, search the web, and use tools.
  • Extensible: Compatible with MCP for adding third-party integrations and allows users to build their own custom work surfaces.