One-page explainer
Manage AI coding work from one place.
aidd gives you one place to plan AI coding work, run the assistant you choose, and review the results. It keeps a backlog and run history for each project, on your own machine.
Pick your background
This page is the general tour. If you're coming from a specific kind of tool, there's a version written for you:
The problem it solves
As work spreads across projects and agent sessions, you need somewhere to track what's planned, what's running, and what still needs review. Conversations are part of that record, alongside feature lists, test results, and decisions.
aidd brings those records into a local control panel. You can launch work, watch a run, review its results, and move between the projects you've added. The coding still happens through the agent or model you choose.
How it works
The coding workflow has four parts:
Plan
Turn a spec into features with acceptance criteria and dependencies.
Build
It does the work in small, tracked steps you can watch live.
Verify
The agent runs the checks required by the project and reports the results.
Record
Keep run logs, feature status, and usage data when the backend reports it.
aidd checks completion reports against feature status and flags blocked work for review. A completed run still needs your judgment about the changes and whether the checks covered enough.
What's inside
One dashboard, all your projects
Projects, running jobs, results, and history live in a single local web panel.
A director for the whole fleet
A built-in "Director" looks across everything you're building and suggests what's worth doing next. You can chat with it, and it can kick off work for you.
Repeatable playbooks
Save multi-step workflows as recipes - "review, fix, verify" - and rerun them on a project or put them on a schedule.
Built-in quality inspections
Built-in audits ask the agent to inspect security, performance, and architecture, then record evidence-backed findings as tasks to review.
Bring your own AI
It works with the assistants and models you already use - ten backends in all, including Claude Code, Codex, Cline and OpenCode, OpenAI-compatible services, and models running privately on your own machine through Ollama or LM Studio.
Steer it from your phone
A Telegram bridge, included with aidd, lets you message the Director from anywhere and have it start supervised work while you're away from the desk - once you connect your own bot and allow-list your chat.
Local records, your choice of backend. aidd stores project metadata, run history, settings, and usage data locally. It has no account or hosted control plane and sends no telemetry. Coding runs send prompts and relevant project content to the backend you choose, which may be a cloud service or a local model. Configured integrations and downloads you initiate can also make network requests.
Who it's for
Anyone using AI to build or maintain software who wants more than a chat window: solo developers running several projects, tinkerers who want their AI work to be repeatable and reviewable, and people on teams experimenting with agent-driven development who need visibility into what the machines actually did. aidd is a single-operator tool - there are no accounts or shared workspaces, so each person runs their own local panel.