02 / HOW IT WORKS

Set the goal.
Approve the plan.
Return to
the result.

DAVE turns an approved outcome into a durable delivery run. Your coding agents build; DAVE schedules, verifies, recovers and returns the correct result.

Request early access CONNECT → APPROVE → LEAVE → TEST
THE ALWAYS-ON LOOPSix stages / Only two human gates
01

Connect

Install an owned runner on Windows, macOS, Linux or a VPS. It works with your repositories and the AI coding agent access you already use.

RUNNER ONLINE
02

Ask

Tell DAVE the outcome in chat, Jira or Cockpit. It gathers the relevant code, tickets, product rules and previous decisions before it proposes the work.

CONTEXT FOUND
03

Approve

Review the tasks, order, risks and definition of done. DAVE cannot approve its own plan or silently expand the scope later.

SCOPE FROZEN
04

Leave

DAVE schedules ready tasks across the right agents, projects and owned runners. Dependencies wait while independent work keeps moving.

DELIVERY RUNNING
05

Recover

If an agent stalls, CI fails or a runner goes offline, DAVE preserves the state, classifies the failure and resumes safely instead of starting from zero.

STATE DURABLE
06

Test

Tests, review, CI and the exact staging build must agree first. You receive one message with the working result and proof, then judge reality.

READY FOR QA
THE OUTPUTThe summary replaces the transcript

Don’t read
the transcript.
Inspect the result.

01GoalBUILD OUTCOME
02Spec hash7F3A
03Tests128/128
04PR #248REVIEWED
05StagingLIVE
06ReadyFOR QA
SERIAL: DAVE-PIPE-A00248SYSTEM: DAVEVERIFIED BY: HUMAN

PRIVATE BETA / EARLY ACCESS

Bring one real feature. See how far it moves without you.

Request early access