Data KineticExplore dkify

Data Kinetic DK-OS

Turn software delivery into a system that proves itself.

DK-OS turns briefs into governed delivery—coordinating repositories, agents, models, budgets, and evidence without replacing the tools your teams already trust.

DK-OS / RUN CONTROLPROJECT: ALCHEMY
GROOM
B031REFUSED
DESIGN
B028QUORUM 2/2
DELIVER
B0243 WAVES
REVIEW
B019CI PASSED
DONE
B012MERGED
B031 refused — “improves UX” is not a measurable outcome
RUN 0247GATES 08/09BUDGET 62%
Built for evidence.Ready for scale.
Git-native truthEvery change stays reviewable
Gates with consequencesProof before progress
Controlled spendBudget checked before dispatch
Agent flexibilityOne method across runtimes

The closed delivery loop

Work moves forward when the evidence does.

From a measurable brief to a reviewed pull request, every transition is observable, governed, and written back as durable context.

PROJECT / DELIVERY SYSTEMLIVE EVIDENCE
B0243 WAVES

Execute the dependency graph

Safe work runs in parallel, contended files serialize, and each task stays behind its gates.

7 tasks · 3 isolated worktrees
RUNNERcodexMODELroutedBUDGETwithin limitRECORDgit + ledger

Orchestration with memory

Make the whole delivery system legible.

DK-OS connects the work already moving through your repositories and turns it into one controlled operating view.

01

Start with intent, not a session

Bring in briefs and work items, preserve external references, and dispatch the right delivery stage without losing the original context.

02

Watch every run become evidence

Durable queues, lifecycle events, streamed progress, verdicts, and failure states make unattended work legible while it is happening.

03

Control a fleet, not another seat

See runs across teams and repositories, attribute model usage, and stop dispatch before an organization crosses its budget.

04

Coordinate beyond one repository

Projects publish status, exchange directed handoffs, and share append-only decisions through a mesh designed for cross-project delivery.

05

Keep every tenant in its lane

Organization-scoped access, row-level isolation, throttling, and immutable audit events protect the boundaries around every run.

06

Let the repository own the truth

The interface projects specs, branches, pull requests, checks, and run records. Rebuild the index without losing the delivery history.

Confidence has to be earned

The refusals are the product.

A useful delivery system does not turn everything green. It makes missing metrics, red gates, unresolved reviews, and human escalations impossible to hide.

See the control model
ATTENTION REQUIRED03 ITEMS
GROOM / B031Metric rejected

“Improve UX” cannot be verified. Define the measurable change.

REFUSED
PROVE / B026Policy gate failed

Read-only enforcement found a write path in the reporting layer.

BLOCKED
REVIEW / B019Human judgement needed

The architecture critic escalated an irreversible boundary change.

ESCALATE

One method, three layers

Open at the core. Controlled at scale.

Adopt the delivery method without an account. Add DK-OS when many repositories, teams, and runs need one operating fabric.

FREE · MIT

01 / THE HARNESS

dkify

The repository-local specification, planning, task, gate, and execution system. It remains useful on its own, with your agents and infrastructure.

ORGANIZE
COMMERCIAL

02 / THE FABRIC

DK-OS

The shared surface for backlog flow, live runs, evidence, budgets, tenancy, fleet health, and cross-repository coordination.

DELIVER
YOUR CONTROL

03 / THE RUNTIME

Your environment

Branches, pull requests, checks, and deployment stay inside the systems and control plane your organization already owns.

The open-source companion

One delivery method. Seven agent runtimes.

dkify installs one canonical command pipeline into the tools your developers already use. The work stays in readable files; gates, dependency order, and run history stay with the repository.

Explore dkify on GitHub
DKIFY 2.0PIPELINE READY

$ dkify init platform --ai codex,claude

  1. 01/dk.specify
  2. 02/dk.clarify
  3. 03/dk.plan
  4. 04/dk.tasks
  5. 05/dk.stage
  6. 06/dk.implement
Claude CodeCodexgooseCursorGemini CLIQwen Codeopencode
MITLicensed harness
06Canonical delivery stages
07Agent runtimes
01Repository-owned truth

Control is part of delivery

Autonomy inside boundaries you can inspect.

DK-OS gives unattended work an organizational perimeter: who owns the run, what it can spend, where it can execute, and what evidence it must return.

  • Organization-scoped isolationAccess and data stay inside the owning tenant.
  • Pre-dispatch budget controlRuns stop before spend crosses the active ceiling.
  • Durable, observable executionQueued work survives restarts and reports lifecycle events.
  • Repository-owned recordSpecs, decisions, branches, checks, and pull requests remain authoritative.

Bring a real delivery constraint

Show us where the method breaks down.

Start with the backlog, gate, handoff, or run you cannot currently see or control. We will map the smallest useful DK-OS operating loop around it.

Prototype only — no data is sent.