# Doodad.ai — Agent Guide

## What the product does

Doodad.ai helps builders who run multiple AI/software projects keep them organized and moving. Core features:

- A dashboard that collects each project's links, GitHub repos, live URLs, screenshots, and AI agents/tools in one place.
- Task lists and project history (including GitHub integration) so users can resume work without losing context.
- AI recommendations ("virtual cofounder") for user acquisition, feature prioritization, and content drafting.
- Public profile pages (`/public/:username`) and public build logs (`/public/buildlog/:username`) that users can share.

## Key routes

| Route | Purpose | Auth required |
| --- | --- | --- |
| `/` | Landing page | No |
| `/pricing` | Plans (Free / Pro / Team), billing FAQ | No |
| `/share` | Create/manage a public profile | Yes |
| `/public/:username` | Public profile page | No |
| `/public/buildlog/:username` | Public build log | No |
| `/dashboard`, `/analytics`, `/settings`, `/onboarding`, `/admin` | Authenticated app | Yes (Clerk) |
| `/api/*` | JSON API for the app | Yes (Bearer token), except public analytics endpoints |
| `/llms.txt`, `/agents.md`, `/robots.txt`, `/sitemap.xml` | Agent/crawler documentation | No |
| `/.well-known/agent-card.json`, `/.well-known/ai-agent.json` | Agent manifests | No |

## How agents should interact

### Safe actions

- Read all public pages: `/`, `/pricing`, `/public/*`.
- Read protocol files: `robots.txt`, `sitemap.xml`, `llms.txt`, this file, and the `.well-known` manifests.
- Answer questions about the product, plans, and limits using the pricing page and its FAQ.

### Actions requiring a signed-in human

- Signing up or signing in (Clerk modals; do not create accounts autonomously).
- Creating, editing, pausing, or deleting projects, tasks, or AI recommendations.
- Starting a Stripe checkout or managing a subscription.

### Prohibited actions

- Never click elements marked `data-agent-danger` (destructive actions such as deleting a project, task, or recommendation). These require explicit human confirmation (`data-agent-confirm`).
- Never complete a purchase or enter payment details on behalf of a user.
- Never attempt to access `/api/*` endpoints without credentials supplied by the user.
- Do not submit forms on authenticated pages.

## DOM hints

- Primary calls to action carry `data-testid` and `data-agent-action` attributes (e.g. `data-agent-action="signup"`, `"view-demo"`, `"select-plan"`).
- Navigation containers are marked with `data-agent-nav`.
- Destructive controls are marked with `data-agent-danger` and `data-agent-confirm`.
- Interactive elements expose a visible `:focus-visible` outline for keyboard-driven agents.

## Canonical docs

- `https://doodad.ai/llms.txt` — concise product summary and link map.
- `https://doodad.ai/.well-known/agent-card.json` — A2A-style capability card.
- `https://doodad.ai/.well-known/ai-agent.json` — guardrails (allowed/disallowed actions, contact).

## Contact

- Support: support@doodad.ai
- Sales: sales@doodad.ai