Rowboat

Rowboat is the rare assistant that improves the longer you leave it alone. It builds a linked note graph from your mail, meetings and chat on your own machine, and the trait that decides it is accumulation: by week six it answers using context you never re-explained. Code mode then hands that context to parallel Claude Code or Codex sessions and drives them with it.

The price is scope. It wants to be your mail client, browser and note taker at once, which is a lot of habit to move. Pick it as one desktop coworker across your work. Pick Claude Code alone when only the repository matters.

The risk is blast radius. The design point is one local index covering mail, meetings, chat and past assistant conversations, and agents can be triggered by an incoming message or by a schedule, so something acts on content you have not read yet while holding access to everything you ever indexed. Untrusted text arriving by mail lands at precisely the wrong layer of that design.

Scope the account and the trigger before enabling anything unattended. Done right: the bundled browser is separate from your everyday one, so the assistant only holds sessions you deliberately signed into there. That is a boundary rather than a promise.

The design choice worth studying is durability of context. Most assistants reconstruct their working set per request by searching transcripts. This one maintains a standing index: sources are written as linked notes with relationships stated rather than inferred, and the meeting recorder captures microphone and speaker, transcribes live, then writes a summary back into the same structure so later work inherits it.

The consequence is traceability. A wrong answer leads back to a specific note, and that note can be corrected by hand, which similarity search does not offer. No evaluation of retrieval quality is published, so the advantage is architectural rather than measured.

Good product, undefined business. The application downloads free, runs entirely on the user's machine, and keeps everything in files the user owns outright, which is excellent for adoption and impossible to meter. Seventeen thousand stars is distribution, and none of it is billable in the current shape.

The obvious next move is a team tier: shared graphs, an administrator, and synchronisation nobody wants to build. That is precisely where the local-first promise gets tested against the need for recurring revenue. Likely acquirer: a productivity suite wanting a memory layer it cannot assemble from mail alone. Position: watch for the first paid tier, and read what it moves off the laptop.

Nothing to buy, and that is the trap. Each installation assembles a complete copy of one person's correspondence, calendar and meeting summaries on that person's device, with nothing central behind it: no single sign-on, no provisioning, no audit trail, and no way to revoke or wipe when somebody leaves. Every user performs their own Google authorisation by hand, which is sixty separate grants nobody is tracking.

There is no pipeline story and no review integration, so this is personal productivity, not a workflow component. Support is a chat server and a repository. Not yet. Revisit if a managed edition arrives with device policy and visibility into what has been indexed.

Apache-2.0, and the configuration is a directory rather than a database. Each optional service gets its own file under ~/.rowboat/config: deepgram.json for voice in, elevenlabs.json for voice out, exa-search.json for research, composio.json for external tools, all sharing one apiKey shape, so the whole thing goes under management and the secrets stay out of the repository. Protocol servers plug in beside them.

Models come from Ollama or LM Studio, so the assistant runs against my hardware and never needs a cloud. The notes are plain Markdown, which means my own scripts read the same graph the agent reads. That is the version of local-first I actually wanted.