Features
Everything you need tosee work before it stalls
Flowmates pairs on-device analysis with signals a team lead can actually act on, so a distributed team gets understood without being watched.
Blocker detection
Sustained effort on a task that is not moving forward is the signal. Flowmates computes it from the six fields the agent publishes, so a stall surfaces while it is still this week’s problem.
Worker-owned architecture
Local inference stays on device under the worker’s control. Exports are explicit, with no silent exfiltration of screenshots, and the source is open to read if you want to verify the behaviour yourself.
Overload signals
Context switching, meeting sprawl and interruption load, surfaced per stream of work. Enough to notice a team member drowning before the retro does.
The Friday report
The week summarized from what actually happened: what moved, what stalled, where the time concentrated. Ninety minutes of Friday afternoon, back.
Focus depth, per stream of work
Deep work blocks, switch counts and recovery windows attached to the work itself. Never a leaderboard, never a comparison between teammates.
Toolchain aware (roadmap)
Upcoming connectors will map local signals to Jira, Linear or GitHub for teams that want ticket hygiene, still driven by worker consent.
How Flowmates works
Install, work normally, and let the week explain itself before the week is over.
Install the agent
Each person installs the desktop agent and enters one team key. No accounts to create, and the heavy models stay on the machine.
Work like you always do
No manual timers, no performative check-ins, no screenshots. Flowmates reads the shape of the work, not the keystrokes.
Talk before Friday
Stalled work surfaces as a fact and an invitation: “TICKET-42, fourteen hours over three days, sixty-eight percent in diagnosis.” Never “Paul is behind.”
Run it on your own team for a week.
Two to four people, one team key, one week. Then decide whether Friday got easier.