Reference
Agent commands
You type /faundr:… commands in your AI agent, inside the folder connected to your project. In Codex, swap the / for a $: $faundr:resume. Each one hands the agent a playbook for a specific job: review security, fix an error, write tests, roll back to an earlier checkpoint.
How to use them
- Type
/faundr:in your agent ($faundrin Codex) to see the list and pick one. - Anything after the command is your request, in your own words:
/faundr:undo before I touched the login. - The agent runs many of these on its own when the situation calls for it. Ask "where did we leave off?" and it runs
/faundr:resume. The Auto? column in the tables below tells you which ones. - The commands only you run are the ones that change something important, or where you decide when they run.
Every dashboard section has a Commands tab with the commands for that area and a ready-to-copy example. This page collects all of them.
Connection
| Command | Auto? | What it does |
|---|---|---|
/faundr:login | no | Connects your agent to your account through the browser. Check the code and click Authorize. Tokens live in My account → Plugin token, where you can delete them. |
/faundr:setup [--refresh] | yes | Sets up Faundr in the project in one pass. It does the safe parts on its own (connected folder, code map, checks, status line, skills) and asks about the rest (.gitignore, AGENTS.md, Supabase token, live app errors). Running it again leaves what's already right alone. |
/faundr:link [name or id] | no | Connects the current folder to a project. With no argument, it lists your projects and asks which one. |
/faundr:status | no | Tells you whether everything is connected and what's left to set up. |
/faundr:statusline [--force] | no | Turns on the Faundr status line at the bottom of your agent's terminal. Claude Code only. |
/faundr:mod [what the mod does] | yes | Builds or adjusts a mod for your agent (side panel, bar above the prompt, notice, command) following best practices, and validates and tests it before handing it over. Claude Code only. See Mods. |
/faundr:agents-md [--file CLAUDE.md] [--off] | no | Copies your rules and decisions into AGENTS.md so agents that don't run the plugin (like Cursor) follow the team's memory too. |
Work and memory
| Command | Auto? | What it does |
|---|---|---|
/faundr:track | yes | The playbook the agent follows to record what it's building on its own: the feature, tasks, decisions and next milestones. You never need to call it. |
/faundr:planning [what you want to build] | yes | Plans before building. It sizes the request, looks at the project, asks a few questions with a recommendation and shows you a summary to approve, including what the agent decided by itself. Once you approve, it saves the feature, the decisions and the definition of done. |
/faundr:feature <title> [— description] | no | Starts a feature, planning it first the way /faundr:planning does, and makes it the current feature for this folder. |
/faundr:inventory | yes | For a project that existed before Faundr. Lists what the app already does, based on the code map, and records each item as a completed feature once you approve. |
/faundr:task <description> | no | Adds a task to the current feature. |
/faundr:done <part of the name or no.> | no | Marks a task as done and shows what's left. |
/faundr:board | no | Shows your features with each checklist, and sums up what's in progress and what comes next. |
/faundr:tidy-board | yes | Cleans up the Features board. It checks every feature in progress and every upcoming milestone against your commits and code, then proposes marking finished work as done (with proof), discarding what won't happen and keeping the rest. It only changes what you approve. |
/faundr:decision <decision> [because …] | no | Records a technical decision. From the next session on, the agent won't contradict it. |
/faundr:rule <rule> [because …] [in <files>] | no | Records a team rule. With files, the agent only sees the rule right before editing those files. If the rule can be checked against the code's text, Faundr warns before any edit that breaks it. |
/faundr:concern <what worries you> [— details] | no | Records a concern (P-n) to review later. It doesn't start working on it. |
/faundr:resolve [P-n] | no | Marks a concern as resolved. With no number, it lists the open ones. |
/faundr:memory-check [rule or area] | yes | Checks whether the code still follows your rules and decisions, and logs each contradiction as a concern. It doesn't change code. |
Continuity
| Command | Auto? | What it does |
|---|---|---|
/faundr:resume | yes | "Where did we leave off?" Covers the last session, what's still pending and what to watch out for. It asks where you want to start and waits for your answer. |
/faundr:handoff [what to record] | no | Writes the handoff note: what's half done, what's left and what to watch out for. |
/faundr:undo [when or what] | yes | Rolls your files back to a checkpoint. It shows you what will change and only rolls back once you approve. The state before the rollback is saved too. |
/faundr:parallel [stages] | yes | Builds several planned stages at the same time. It takes the ones that can start now (the previous stage is done), opens a workbench for each, sends one helper to build each stage there, then merges them into the main project one at a time, with the tests. It only starts and merges once you approve. |
/faundr:merge [what was done] | yes | Merges the work from your workbench into the main project. It pulls the latest version, merges, runs the tests, explains any conflicts and pushes. It only merges once you approve. |
/faundr:inbox | yes | Fetches the requests, messages and notes your team left for you in the dashboard, and handles each one. Requests from someone else only go ahead once you approve. Run /loop 5m /faundr:inbox to have it check every few minutes. |
Overview, Stack and code map
| Command | Auto? | What it does |
|---|---|---|
/faundr:overview | yes | Writes or updates the project Overview. |
/faundr:stack | yes | Writes or updates the Stack. Tell the agent anything the files don't say. |
/faundr:map | no | Builds the code map for the first time, or rebuilds it now. |
/faundr:map <question> | no | Asks the map a question ("how does login work?", "what breaks if I change X?", "how does A reach B?"). The agent answers with file and line references. It and its subagents already query the map on their own from the terminal. |
Design
| Command | Auto? | What it does |
|---|---|---|
/faundr:design [--files <files>] | yes | Without a design.md, it creates the guide and asks what you want. With one, it audits every screen, measuring in the browser if the dev server is running. With --files, it runs a quick review of just those files. |
/faundr:design-fix [D-n | --all] | no | Fixes design issues following design.md. With no number, it works through them in order of importance and pauses after each stage. With --all, it fixes everything without stopping. |
/faundr:design-report <what you saw> | no | Logs an issue you spotted on screen. It becomes a D-n. |
/faundr:design-showcase [component] | yes | Draws the component replicas in the Showcase. With no argument, it redraws only the outdated ones and adds the missing ones. |
Security, launch and privacy
| Command | Auto? | What it does |
|---|---|---|
/faundr:security-scan | no | Runs the non-AI check now and summarizes what it found. |
/faundr:security [--files <files> | S-n] | yes | AI security review: who can access what. Every issue goes through an independent verifier before it's logged. Anything that depends on configuration outside the code becomes "Needs check". With S-n, it rechecks a "Needs check" item. It doesn't fix anything unless you ask. |
/faundr:security-fix [S-n | package | --all] | no | Fixes one issue, or every serious one (--all fixes all of them without stopping). An issue only closes once the check stops finding it. For a leaked key, it walks you through rotating it with the provider. |
/faundr:launch [--url <site>] | yes | "Ready to launch?" Builds your code, runs the tests and the security check, tests your database as a visitor would, checks the live site and ends with Can launch: YES or NO, starting with what blocks launch. |
/faundr:privacy [policy | terms | cookies | deletion] | yes | Works out which privacy laws apply (LGPD, GDPR, CCPA/CPRA) from where your users are, shows the personal data your app collects and what those laws still require, and writes the drafts once you approve. |
Code quality
| Command | Auto? | What it does |
|---|---|---|
/faundr:quality-scan | no | Runs the non-AI check now and summarizes the results by type. |
/faundr:quality [--full | --files <files>] | yes | AI review of what changed, or of the whole project. It only reports issues and changes nothing. |
/faundr:quality-fix [Q-n | --all] | no | Fixes issues without changing what the app does (--all fixes every open one without stopping). It asks before touching many files or removing packages. |
Tests and errors
| Command | Auto? | What it does |
|---|---|---|
/faundr:tests [feature, file or T-n] | yes | Asks what can't break, links features to the tests that protect them and writes the missing tests, watching each one fail before trusting it. It never installs anything without your OK. |
/faundr:tests-fix [test or file | --all] | no | Fixes failing or flaky tests without deleting, skipping or loosening them. --all fixes all of them without stopping. |
/faundr:error-fix [E-n | --all] | no | Fixes an error and only marks it resolved once the same command passes again. It starts from the recipe if this error (or the same one) was fixed before in the project, and saves the recipe at the end. With no number, it starts with errors that came back. --all fixes all of them without stopping. |
/faundr:errors-setup | no | Connects your live app to Faundr so errors your users hit show up in the dashboard. |
/faundr:browse <address> <goal> | yes | A fast browser. It runs a task on a website in seconds (click, fill in, search, open screens), with no screenshots and without spending your plan's tokens on every click. Each step is picked by Jev, an outside service that needs a key (faundr browse-key). Use it to test your app's flows: login, sign-up, navigation after a deploy. Passwords never go in the command; they stay in an environment variable. When the last step would pay, buy, send, delete or publish, it stops before that button. |
Skills and workflows
| Command | Auto? | What it does |
|---|---|---|
/faundr:skills [skill name] | yes | Connects the skills you already use (from the repo, your personal ones and the ones from enabled plugins) to Faundr. The agent writes when to use each one and its topics. They show up under Workflows → Skills and in the canvas palette, and the agent gets a reminder to use one when a request touches its topic. It also marks the skills that find problems (security, tests, quality, errors) as Connected, so what they find shows up in the dashboard. See Workflows. |
/faundr:workflow [what the flow does] | yes | Drafts a project workflow: how your team wants the agent to handle a type of request ("whenever it's about design, do this, then that"). It saves a draft, which only takes effect once you approve it in the dashboard under Workflows, where you can also edit the steps on the canvas. |