Documentation · early preview
Make agent work repeatable—and provable.
Doku is the persistent work layer for AI agents. Start with the model, then author a Playbook that defines the result, the work, and the evidence required to finish.
PLAYBOOK / LAUNCH REVIEWRUNNABLE
Run the work3 milestones · 7 tasks01
Prove the claimsEvidence stays attached02
Gate completionHuman approval where needed03
THE OPERATING MODEL
A straight line from intent to verified work.
01PlaybookDefine the outcome and proof.
02RunGive an agent bounded work.
03EvidenceRecord what made the claims true.
04ApprovalKeep consequential decisions human.
INSIDE DOKU
The work stays visible—from setup to proof.
Connect the client where your agent works, then select a governed Playbook with a published outcome, required access, and an explicit workflow.
START HERE
Learn the system in three reads.
These pages are the smallest useful path from first principles to a valid package.
What is Doku?
Understand authority, Playbooks, Runs, evidence, and approvals.
02 · AUTHORAuthor a Playbook
Build a package with Outcomes, Claims, Variables, Tasks, and Milestones.
03 · EXECUTEAgent runtime guide
Claim a Run, work through Doku-issued Work Items, and finish safely.