Project dashboard
Code map
The Code map is an index of your project: its files, its functions and what uses what. Faundr's engine builds it on your computer, without AI, from your code and docs.
It's useful in two ways:
- For the agent: instead of opening dozens of files to understand part of the code, it asks the map and gets back only the snippet that matters. That uses fewer tokens and gives more accurate answers. Insights shows how much this saved.
- For you: see how the project is organized, which parts talk to each other and what could break if one of them changes.
Building the map
The map builds itself in the background during your agent's first conversation in the connected folder. To build or rebuild it right now, run:
/faundr:map
From then on, the map keeps itself up to date. When the agent edits files, the map is rebuilt at the end of the response. Queries also refresh any changed files before answering.
What's left out
The map only reads your project's code and docs. It skips:
- anything
.gitignorealready ignores, and folders likenode_modules,distand.venv; - the project copies your agent creates for subagents (
.claude/worktrees); - minified libraries and generated code (
*.min.js, build bundles, files with enormous lines); - files with passwords and keys, such as
.env; - anything listed in
.faundrignore(see below).
In a project with more than 5,000 files, code goes in first and Faundr tells you what was left out.
Keep folders out of Faundr with .faundrignore
To leave a folder out on purpose (third-party code, a repo you cloned just for reference, generated files), create a .faundrignore file at the project root. It uses the same format as .gitignore: one folder or file per line. For example:
refs/ vendor/
.faundrignore isn't just for the map. The folder is also left out of Off-plan in Architecture, the design rules and the checks that look for your project's parts (subfolders with their own dependencies, like frontend/ or apps/web/).
Languages
The map understands JavaScript and TypeScript (including React Native, Vue, Angular, Svelte and Astro), Python, Flutter and Dart, PHP, Go, Java, Kotlin, C#, Rust, C and C++. HTML, CSS, Tailwind and Docker have no functions to map, so they stay off the map (Faundr's other checks still cover them). Markdown docs are included in every project.
To see what Faundr does in each language, open Stack → What Faundr covers in the dashboard, or the Supported languages page.
Tabs
| Tab | What it shows |
|---|---|
| Overview | The project split into areas, arranged in columns: screens, UI pieces, entry points (APIs and plugins), logic, data, docs and technical. Click an area to see what's in it and which other areas it talks to. Click again to open it. |
| Files | Your files, each with an icon for its type. Click a file to see what it uses and what it's used by, and to jump to its code. |
| Code | The full technical map: every function and file as a node, grouped into communities. For when you want the details. |
| Report | The text report generated by the engine, with the communities and the central pieces. |
See how two parts connect
In Overview and Files, every connection has Show how they connect, which explains in two sentences, without AI, how one part uses the other. The Explain button copies a ready-made request for the agent to explain it in detail (/faundr:map).
Options at the top
- Technical details: in Overview and Files, also shows configuration, tests and external packages, which are hidden by default.
- Hide names: in the Code tab, hides the names and the side panel. Use it to show the map (in a pitch or a post) without exposing your code.
Asking the map
Ask with /faundr:map <question>, or let the agent do it on its own. Faundr tells the agent, and any subagents it opens, to query the map before reading files.
Questions that work well:
/faundr:map how does login work? explain in plain language/faundr:map what breaks if I change AuthGate?/faundr:map how does the projects screen reach the database?
.faundr/mapa/, inside your project.Commands
The Code map is built by Faundr's engine, without AI, from the code and the docs. Run these commands in your AI agent, from the folder linked to the project.
After the first run, the map updates itself whenever the agent edits files.
/faundr:mapBuilds the code map for the first time (or rebuilds it now) and sends it to the dashboard.
/faundr:map <question>Asks a question about the code. The map returns the most relevant parts with their code and who uses them, and the agent answers citing file and line. Also works for "what breaks if I change X?" and "how does A reach B?".
e.g. /faundr:map what breaks if I change AuthGate?
!faundr graph-query "<question>"The same question, straight in the terminal and without AI: the map returns the most relevant parts with their code.
e.g. !faundr graph-query "how does login work"
!faundr graph-path "A" "B"Shows how two parts of the project connect.
e.g. !faundr graph-path "LoginPage" "supabase"
!faundr graph-callers "X" [--out | --both] [--depth 2]Who uses X. --out shows the opposite (what X uses); --both, both. --depth 2 shows the chain effect.
e.g. !faundr graph-callers "AuthGate" --depth 2
!faundr graph-skeleton <file>A summary of a file: each function with its header, without the body.
e.g. !faundr graph-skeleton src/lib/auth.ts
!faundr graph-grep "<text>" [--in <folder>] [-i]Searches for a text and groups the results by the function where they appear, most used first. -i ignores case.
e.g. !faundr graph-grep "session" --in src