How to use DDesk

What DDesk is

DDesk is an AI software delivery desk. You connect a repository (or create a greenfield app), open a story, and specialized AI employees run product → architecture (Spec Kit plan) → development → QA → review. Humans approve acceptance criteria, pull requests, and production. Spec Kit is used for planning only — your AI Developer employees implement the work.

First 15 minutes

1. Sign in or create an account (email, Google, or GitHub).
2. Open Dashboard and follow Get started, or open guided setup.
3. Go to Projects — Connect GitHub (repo and branch), or paste an HTTPS URL, or use Create standalone app for a scaffold.
4. Optional but recommended: on the project page, set Spec Kit constitution (principles, coding/security standards, new vs existing app mode). Leave blank to use defaults.
5. Create work via Support (Auto-fix), Requirements → suggest stories, or Stories directly — then Start delivery.
6. When a gate is waiting, open Approvals. On the fast lane, acceptance criteria and a pull request with green checks can move on without a click. Production stays with you.

Projects & workspace

The sidebar project selector sets the active workspace for Stories, Requirements, Health, and Operations. Switch projects anytime; lists and pipelines filter to the selected project.

• Existing repo — paste the HTTPS clone URL and default branch so sandboxes can clone safely. Add a Git credential on the project (or org) if needed.
• Create standalone app — with an org GitHub token, DDesk can create a private repo, seed a template, and open a first requirement. Without a token, point at an empty repo and seed manually.
• Spec Kit constitution — standards Architect uses when planning after AC approval. Development cannot start until the technical plan has clear ordered tasks. Implementation still runs through your AI employees.
• Site URLs — set preview / staging / production on the project for Analytics and deploy flows.

Create work (three paths)

Support (fastest first win)

Open Help & support, file a ticket against your project, and turn Auto-fix on. DDesk creates a requirement and starts the employee pipeline. Use this for bugs and small fixes you want the desk to own end-to-end.

Requirements → suggest stories

On Requirements, write a one-line or detailed need, then Suggest stories. With an OpenAI or Anthropic key in Settings → Org secrets (or OPENAI_API_KEY on the server), an LLM Product Owner expands any one-liner into epics, stories, and AC; otherwise domain packs (e.g. e-commerce) still apply. Select what you want, create drafts, then start delivery.

Stories directly

On Stories, create a story or pick a template. From the board or story detail, use Start delivery to kick off the pipeline. Artifacts (plans, tests, reviews) accumulate on the story; Code Changes and Pull Requests show engineering output.

Delivery pipeline

After you start delivery, AI employees advance the run until a human gate or completion:

1. Product Owner — drafts / revises the story and acceptance criteria.
2. You approve AC — on the full lane this is a human gate. On the fast lane it is recorded automatically when acceptance criteria exist.
3. Architect — Spec Kit–style plan (constitution, clarify, checklist, research, contracts, ordered tasks). No coding yet.
4. Developer — AI employee implements the ordered tasks in a sandbox.
5. QA → Tech Lead — tests and PR. Failed QA can send work back to Development.
6. You approve the PR (and production, on the full delivery lane).

Most work uses the fast lane (skips senior review and DevOps). Ask your workspace admin if you need the full lane with staging and production gates.

Watch progress on the story page, Workflows, Tasks, Code Changes, and QA.

Approvals & gates

Open Approvals when the badge shows pending items. Typical gates:

• Story acceptance criteria — edit if needed, then approve, request changes, or reject.
• Pull request — you approve merge intent, unless the fast lane already accepted green checks.
• Production deploy — on the full lane only; staging may run first.

Rejecting sends work back into the desk with your notes. When two or more story gates are waiting, select them and approve the page together. Production is always a separate decision.

Health, analytics, billing

• Scans and Problems — code-health packs; fix actions can open delivery through the same approval path.
• Analytics (Starter+) — site quality when a project preview URL is configured.
• Report — summary for the active project.
• Billing — plan limits, seats, monthly support tickets, and Manage in Stripe after checkout.
• Costs — AI usage for your projects.
• Employees — which AI roles exist on the desk.

Settings you may use as an operator

Under Settings: organization name, budgets, secrets, and (if you integrate tools) machine API keys for your own scripts. Coding engine keys are usually configured by your admin on the server — ask them if Developer steps stay on the stub engine.

Help & legal

Account questions: Help & support. Product bugs on a connected repo: sidebar Bug reports. Legal: Privacy · Terms · hello@dedicinfotech.tech