Skip to main content
Rowboat Desktop provides AI assistance that’s grounded in your work context. From drafting emails to generating presentations, every feature is designed to work with memory that compounds over time.

​
Core Features

​
AI Copilot

Rowboat Copilot is your AI assistant for everyday work. It combines thorough explanations with genuine enthusiasm and helps with:
  • Email drafting - Respond to emails with full context from past conversations
  • Meeting prep - Get briefed on attendees, past decisions, and open questions
  • Document collaboration - Create and edit notes in your knowledge graph
  • Presentation creation - Generate PDF slide decks from your work context
  • Project tracking - Summarize project status using accumulated memory
The Copilot has access to your entire knowledge graph and pulls relevant context automatically. No need to search or provide background - it already knows.

​
Knowledge Graph

Your work memory lives in an Obsidian-compatible vault stored at ~/.rowboat/knowledge/.

​
Graph Structure

The knowledge graph is organized into four categories:
People/ - Individual contacts and relationshipsEach person note tracks:
  • Role and organization
  • Past interactions and decisions
  • Commitments made by both sides
  • Open questions and concerns
  • Communication style and preferences
Example: knowledge/People/Sarah Chen.md
Obsidian-style backlinks connect notes together. For example, a meeting with Sarah Chen will link to [[Sarah Chen]] and [[Acme Corp]], creating a web of related context.

​
How the Graph Updates

The knowledge graph updates automatically:
  1. Email sync - Gmail threads are processed and entities extracted
  2. Meeting sync - Transcripts from Google Meet, Granola, or Fireflies are analyzed
  3. Note creation - New people, projects, orgs, and topics get dedicated notes
  4. Context enrichment - Existing notes are updated with new decisions and commitments
1

Sources are synced

Emails and meeting transcripts are saved as markdown in ~/.rowboat/sources/
2

Graph builder runs

The note creation agent processes new/changed source files in batches
3

Entities extracted

People, organizations, projects, and topics are identified
4

Notes created or updated

New notes are created or existing notes enriched with new context

​
Note Creation Strictness

Rowboat supports three strictness levels that control how aggressively notes are created from emails. Meetings always create notes at all levels.
Philosophy: “Meetings create notes. Emails enrich them.”
  • Emails never create new notes (only meetings do)
  • Emails can only update existing notes for people you’ve met
  • Exception: Warm intros from known contacts can create notes
  • Best for: Users with high email volume who want minimal noise
Filters out:
  • All consumer services (Amazon, banks, etc.)
  • Cold sales and recruiting emails
  • Support representatives
  • Generic business pitches
On first run, Rowboat analyzes your emails and recommends a strictness level based on volume and patterns. You can change this setting anytime in ~/.rowboat/config/note_creation.json

​
Background Agents

Rowboat can spin up background agents to handle repeatable work automatically. Examples:
  • Draft email replies in the background (grounded in past context)
  • Generate a daily voice note each morning (agenda, priorities, meetings)
  • Create recurring project updates from latest emails/notes
  • Keep your knowledge graph up to date as new information arrives
You control what runs, when it runs, and what gets written to your local Markdown vault. Agent schedules are configured in ~/.rowboat/config/agent_schedule.json

​
Voice Notes

Record voice memos that automatically:
  • Transcribe with Deepgram
  • Extract key takeaways and action items
  • Update relevant notes in your knowledge graph
  • Link to people, projects, and topics mentioned
Voice notes require a Deepgram API key. Add it to ~/.rowboat/config/deepgram.json

​
Skills System

Rowboat uses a skills catalog to load specialized guidance for different tasks:
  • draft-emails - Structure for drafting email responses
  • meeting-prep - Guidance for creating meeting briefs
  • create-presentations - Generate PDF presentations
  • doc-collab - Create and edit documents in the knowledge base
  • slack - Send messages and interact with Slack
  • web-search - Use Brave or Exa for research
  • mcp-integration - Discover and execute MCP tools
When you ask Rowboat to perform a task, it loads the appropriate skill first to ensure it follows best practices.

​
MCP Tool Support

Rowboat supports the Model Context Protocol (MCP) to connect external tools and services. Available integrations:
  • Web search (Exa, Brave)
  • Slack (messages, channels, users)
  • Twitter/X (posts, mentions)
  • ElevenLabs (text-to-speech)
  • Linear/Jira (issues, projects)
  • GitHub (repos, PRs, issues)
  • Databases and custom APIs
MCP servers are configured in ~/.rowboat/config/mcp.json. Use the Settings dialog in the app to add new servers.

​
UI Features

​
Chat Interface

The main chat interface includes:
  • Message history - Full conversation context
  • @mentions - Reference people, projects, or topics from your graph
  • File attachments - Upload documents, images, PDFs for context
  • Code blocks - Syntax-highlighted code in responses
  • Markdown rendering - Formatted text, lists, and tables

​
Settings Dialog

Access all configuration from the settings dialog:
Configure your AI provider and models:
  • Select provider (OpenAI, Anthropic, Google, Ollama, etc.)
  • Enter API key (for cloud providers)
  • Set base URL (for local/custom endpoints)
  • Choose assistant model
  • Choose knowledge graph model (can differ from assistant)
  • Test connection before saving

​
Version History

Rowboat automatically tracks changes to your knowledge graph:
  • Git-based version control for all notes
  • Automatic commits when notes are created or updated
  • Full history available via standard git commands
  • Stored in ~/.rowboat/.git/
You can use any git client to view history, compare versions, or roll back changes to your knowledge graph.

​
Security Features

​
Command Approval Flow

Rowboat has two types of operations: Builtin tools (no approval required):
  • Reading/writing files in ~/.rowboat/
  • Searching the knowledge graph
  • Parsing documents
  • Managing MCP servers
Shell commands (approval flow):
  • Pre-approved commands in ~/.rowboat/config/security.json run immediately
  • Other commands require one-time user approval
  • Commands that modify files outside ~/.rowboat/ always ask for confirmation

​
Data Privacy

  • All data stored locally - Nothing sent to cloud storage
  • API keys encrypted - Provider keys stored securely on disk
  • No telemetry - Rowboat doesn’t phone home
  • Inspectable - All notes and configs are plain text

​
Next Steps

Configure AI Models

Set up OpenAI, Anthropic, Google, Ollama, or local models

Explore Workspace

Understand the ~/.rowboat/ directory structure

Build docs developers (and LLMs) love

Get started for freeTalk to us