Ship reliably and grow the team by keeping every decision made out loud

Design reviews, incident calls, 1:1s: the decisions your team runs on are made in rooms and then live only in your head, until they are owed back as an ADR, a postmortem timeline or a promotion packet. Earkeep is an always-on transcriber that keeps every one of those rooms on your own device, and the agents you already run against code can read them the same way.

No bot · No cloud · No record button

14-day full trial, no account required.

The meetings that decide it

Design review

What has to survive it is the decision and the reasoning under it, including who argued the other side, because that is the ADR nobody writes in the twenty minutes after the room empties. Six months later the same failure mode resurfaces and the reasoning is gone.

With Earkeep: nothing joins the call and nothing is uploaded, which matters when the room is full of unreleased design detail that cannot go through a vendor cloud. Audio is transcribed on your device and never written to disk, so the confidentiality objection to a notetaker does not apply. How always-on works →

Incident call

What has to survive it is an accurate ordered account of what happened and who said what, before memory smooths it into a story. Nobody starts a recording when they get paged, and the postmortem's whole credibility rests on being right about the sequence.

With Earkeep: capture is already running when the page arrives, so you mark the span once the incident is over and the timeline is already there. The bridge, the follow-up huddle and the retro three days later are all on the same searchable record. From what was said to what you know →

Weekly 1:1s

What has to survive it is the specific dated thing someone did, because a promotion packet built at review time from memory is a packet made of the last three weeks. You do not remember the quiet refactor in March that prevented three incidents.

With Earkeep: connect a calendar and each recurring 1:1 titles its own span, so a year of them accumulates without anyone maintaining a spreadsheet. Transcripts are plain dated files on your own disk, which is what makes a year of them queryable at all. How calendar spans work →

The rest of the week

Those three are where the team's direction is set. They are not where the week goes: Clockwise's analysis of 1.5 million meetings across more than 80,000 engineers put engineering managers at 17.9 hours a week in meetings, about seven more than an IC, costing over nine hours of weekly focus time. Bain found managers giving 50% or more of their time to meetings while two-thirds of those meetings end before a decision is reached, and Jellyfish reported 65% of engineering leaders and their teams burning out in a year.

MeetingYouWhat has to survive it
Daily standupRunWho owns what and what is blocked, without pinging each engineer for it
Sprint planningRunCapacity decisions, and what was committed to rather than discussed
Backlog refinementRun or attendEstimates and scope decisions the tickets are supposed to match
RetroRunAction items that otherwise die between one retro and the next
Cross-team syncAttendAPI contract and dependency decisions both teams must remember the same way
1:1 with your own managerAttendStatus and escalations passed upward accurately
Hiring interview and debriefRun or attendThe hire or no-hire signal with its specific reasoning, before candidates blur
The ad-hoc desk huddleAttendThe decision made informally that nobody calendared or wrote down

Earkeep does not ask you to choose between them. Every recurring row above starts, ends and titles its own span from your calendar, so a twelve-meeting day ends as twelve named transcripts with nothing left to file.

Writing them up is a separate job from keeping them, and the templates are free: templates for all five sprint ceremonies, including the one-experiment rule that stops a retro producing five actions and none, and the 1:1 template for the weekly ones.

Two of these have a page of their own. The incident review is a debrief, and the four questions plus the evidence for them are there. The handoffs that break between your team and another are in cross-functional collaboration, including the study that measured how many dependencies had no conversation attached.

Pricing: one payment, no subscription

€39

once, for one device, not a subscription.

No tiers, no per-seat billing, no recurring bill. Try it free for 14 days with full functionality, no email and no account required. See full pricing and policy details →

FAQ for engineering leads

Can I use Earkeep for architecture reviews and incident postmortems without a confidentiality problem?

Audio is processed only in memory and never written to disk, never uploaded to any API, with no account or cloud involved. That's a stronger position than the bot-based notetakers many engineering orgs already restrict for exactly this reason.

Does this replace the standup or the 1:1?

No. Earkeep doesn't change how you run meetings, it just captures them. You still run the standup the way you run it today; the difference is that what was said in it now exists as a transcript you or your agent can query afterward.

Will my team have to change tools?

No. Earkeep sits underneath your existing stack (Jira, GitHub, Notion, Slack); it doesn't replace any of it. It fills the one gap none of those tools cover: what was actually said in the room.

Can I point Claude Code at meetings I didn't personally attend?

Yes, as long as Earkeep captured the span. Any transcript on disk is queryable by whichever agent you run, in-app or via MCP.

Does the agent's model see my meeting content over the network?

Earkeep's own part stays local: transcripts never leave your device via Earkeep. Once you point an agent at a transcript, whether the content then leaves the device depends on which agent you chose and how it's configured, since most external agents call their own model provider's API. See agents and MCP for the full explanation.

Does always-on capture burn through battery or CPU during a long meeting day?

A voice-activity-detection pipeline only runs transcription when someone is actually speaking, on the same local models other transcription apps use. Always-on doesn't mean always-transcribing.

Stop being the single point of memory

14-day full trial, no account required.