I use aidd differently depending on what I’m starting with. Sometimes I have an idea and no directory. Sometimes I have a repository to get familiar with. With the fleet, I may need help choosing which project needs attention first.

I’ve recorded a run for each of those cases. Pick the one closest to your situation below; the common installation steps are at the bottom of this post.

The fork

Create Fresh begins with a product sentence. It can create a repository foundation, turn the idea into a dependency-ordered backlog, validate both, and stop before the first product feature.

Ingest Existing begins with source code. The normal project-intake recipe invokes codebase-analysis as its first nested step, so I do not run analysis separately before onboarding. Intake continues through its interview, project profile, artifact checks, feature and testing coverage, audits, and a final parking step.

Director begins above any one repository. It snapshots the discovered catalog, ranks evidence, and leaves proposed work pending by default. That is always what happens when I start the cycle myself, including with Run now. In v3 I can separately opt an automatic scheduled cycle into launching a bounded set of eligible suggestions; rank, risk, recipe, project state, and per-cycle limits keep that lane narrow. I use Director when choosing the repository first would already be a guess.

Each route also works on its own if you arrive from the archive.

What do you do?

The common setup

I run aidd locally. From a source checkout, the short route to its control panel is:

git clone https://github.com/NomadicDaddy/aidd.git
Set-Location .\aidd
bun install
bun run start:web

That opens http://127.0.0.1:3210. To pin a version instead of following main, download the source archive from the latest release, extract it, and run the same bun install and bun run start:web commands inside it. As of September 7, 2026 that is v3.0.0, the first release on the reset history and the only tag the repository carries. Either route needs Bun 1.4.2 or newer, because aidd ships no prebuilt executables and the release carries source archives and nothing else.

aidd still needs an agent backend. It can call an OpenAI-compatible provider itself or drive an authenticated agent CLI. I used Codex for the project runs here. I had already signed it in and saved the backend, model, and reasoning level in Settings, so the launch panels inherited those choices.

The other shared setting is the filesystem boundary. My configured project root is D:\applications. Create Fresh may create a directory there, Ingest Existing may discover a clone without .aidd metadata, and Director may include managed projects under that root in its fleet snapshot.