The meeting is over. The tasks are not written.

Paste the notes and get back a short list of real issues — each with an owner, an outcome and the line it came from. File them to Linear or Notion one approval at a time, or keep the list and file nothing.

Turn notes into tasksFree to try. Nothing to install.

How it works

Three steps, and none of them is “configure”.

01
Give it the notes
Paste them, drop a file in, or upload the recording and let it transcribe. It reads all of it before extracting anything — a decision reversed on the last line is still reversed.
02
It writes the list down first
tasks.md appears in the pane beside the conversation: the proposed issues, what it could not tell, and what it deliberately left out. Argue with it there, before anything is filed.
03
You approve each issue
Filing to a tracker pauses and asks, once per issue, showing exactly what is about to be created. Say no and it moves on rather than trying again.

What it can do

It tells a task from a note
“Discuss pricing” is a note. “Decide whether Pro includes seats — Priya, before the board deck” is a task. Everything else becomes context in the body rather than its own issue.
It will not invent an owner or a date
If nobody was named, the issue says unassigned. If no date was given, there is no date. An invented deadline is worse than a missing one, because it looks like a decision somebody made.
It reports what did not resolve
Two people contradicting each other is the most useful thing in most meetings. It quotes both rather than picking one and filing it as fact.

How it is set up

The mechanics, so you know what you are getting before you sign in.

Environment: A Linux container with the notes in it. Nothing is executed — this agent reads and writes text.
Tools: readFile, writeFile, editFile, glob, viewDocument, transcribe. Plus Linear or Notion over MCP, if you connect one.
Model: Any chat model in the catalogue. The work is judgement about what counts as a task, so a stronger model shows up as a shorter list.
Starting files: Seeded with a genuinely messy set of notes — a contradiction, a reversal, an item with no owner, one already done.
What persists: Approvals live in the session, not the tab. Close it mid-review and the pending issues are still waiting when you come back.

Things people ask it

“Turn these notes into issues — flag anything that was left unresolved.”
“What was actually decided here, and what only sounded decided?”
“Transcribe this recording and pull out the action items with owners.”
“File the first three as Linear issues, and leave the rest in the file.”

What it will not do

Every one of these is a real constraint we have hit, not a roadmap item.

It files only what you approve, one issue at a time. There is no bulk approve, on purpose.
It reads what the notes say. If a decision was never stated out loud, it will report the gap rather than close it.
Linear and Notion are optional and connected in Settings. Without one, the deliverable is the file.

Which model does this best

Measured over real runs of this agent, per model. A run counts as finished when it produced the answer on its own — nothing failed, nobody was asked to approve anything, and it did not run out of steps. Read all three columns together: a model that finishes fast by giving up scores badly on the first, and one that completes everything by grinding scores badly on the second.

Model	Completion rate	Median cost	Median time	Runs
claude-sonnet-5anthropic	71%	$0.051	47s	7

Completion rate: Runs that finished the job without you having to step in.
Median cost: What a typical run costs, in credits.
Median time: Wall clock, from the first message to the answer.

Medians over the runs behind each row. A model appears once it has 3 runs on this agent, and the run count is shown so you can judge how much a figure rests on.