closed beta

All your agents, one deck.Hand off the work, hold the overview.

Run many terminal coding agents in parallel, each in its own lane, all reading one shared project brain — so work hands off cleanly between people and agents, and you always see which lane needs you.

Email only. One note when it ships.join the closed beta

  • mac, windows, linux
  • local-first
  • no telemetry

the deck

Run ten agents. Lose track of none.

See every agent at a glance. Lanes that need you float to the top, so nothing waits in a terminal you forgot to check.

live lane deck · focus, groups and board views · drag-and-drop grouping · cross-project dashboard · ⌘K command palette

  • idle
  • running
  • needs you
  • review
  • merged
  • blocked

isolated lanes

Every task in its own lane: worktree, branch, terminal, agent.

A lane is a real git worktree on its own branch with its own working dir, so parallel work never collides. Each gets a stable dev port and lane env vars, and reopening a lane resumes the agent where it left off.

git worktree + branch per lane · new repo, worktree, in-place or folder · per-lane port 3000–3999 · claude --continue on reopen · diff, merge, archive, remove

earned attention

Notifications are earned. Nothing else shouts.

A lane only lights up when its output signals for you — an OSC 9/777 notification or a terminal bell, caught even across split output chunks. You get one quiet, focus-aware notification per hand-off, not a stream of noise.

watches OSC 9/777 and the bell · optional Claude bell injection · one focus-aware OS notification · a dock badge counting lanes · walks you through the ones that need you

shared brain

One shared brain, seeded into every lane.

Every lane's agent reads the same brief. Write it once as markdown and Lanedeck seeds it into each lane, so hand-offs carry the context instead of losing it.

eight knowledge kinds as markdown · seeded into .lanedeck/CONTEXT.md · a managed block in CLAUDE.md · pull and push the team brain · boards live-sync as agents work

built for agents

The deck runs itself.

A private per-user control socket and the lanedeck CLI let an in-lane agent call back into the app — flag its own lane, drop board tasks, write handovers. The seeded CLAUDE.md teaches every agent to use them without being asked.

per-user control socket · lanedeck notify from inside a lane · board, handover and knowledge writes · deck scripting to list and create lanes · seeded CLAUDE.md teaches every agent

never lose work

Reboot-proof. Unattended-proof.

Close the lid, reboot, or hit a usage limit and the work comes back. Projects, lanes, and terminal history reconcile to disk, and armed lanes resume on their own once usage frees up.

session restore reconciled to disk · scrollback restore per lane · auto-resume after a limit reset · usage gauges read locally · no babysitting a blocked agent

0 tokens
to read your usage
every lane
back after a reboot
auto-resume
once usage frees up

local or team

Local-first, or your whole team.

Work fully local as a guest, or sign in to sync the brain with your team. Orgs, invites, and cloud projects when you want them, nothing running when you don't.

fully local as a guest · orgs with owner/member roles · email invites · public-to-org or private projects · read-only export share tokens

  • mac, windows, linux
  • macOS signed + notarized
  • auto-updates
  • no telemetry

waitlist

Join the waitlist.

Lanedeck is in closed beta. Add your email and we'll let you in as room opens up.

join the closed betaemail only, one note when it ships