Faundr, built into your coding agent.
Two commands to install. From then on, your AI follows your project’s rules, remembers where it left off and explains every error in plain language. Ready-made commands are one copy away, and you always know what’s going on, even if you don’t code.
Paste into Claude Code to install
- 1
- 2
Then create your project in the dashboard. It connects your project folder and walks you through the rest.
Eight new things your agent can do. All of it works without any setup.
The plugin runs behind every chat and syncs everything to your Faundr dashboard, so you can see it all and stay in control.
Copy, paste and you’re done. There’s nothing to memorize.
Every area of the dashboard has a Commands tab: the command with a copy button, a one-line description and an example of how to use it. Errors, failing tests and code issues come with a button that copies the fix prompt for you.
Run these in your coding agent, from your project folder.
/faundr:rule <rule> [because …] [in <files>]
Saves a team rule. Name files or folders and the AI only sees the rule when it’s about to edit one of them.
e.g. /faundr:rule Payments only through Stripe because we never store cards in src/payments
/faundr:decision <decision> [because …]
Saves a technical decision to the current feature. It shows up on the feature page and in Memory → Decisions.
e.g. /faundr:decision Use Supabase for login because we already use its database
/faundr:map <question>
Ask anything about your code and get the answer with file and line number. Great for “what breaks if I change X?”
e.g. /faundr:map what breaks if I change the shipping calculation?
!faundr statusline-install
Turns on the Faundr status line on this machine.
Starts with / or !
/faundr:… asks the AI to do the work.
!faundr … runs right in the terminal and answers instantly.
Or just ask in your own words. “Draw the project’s architecture” works too.
Code map · Explain
Checkout calls 2 functions in Payments (createSession and confirm).
Click the line between two parts of your app to see how they connect, in one sentence. Need more? The button copies a prompt asking the AI for the full explanation.
Every problem, explained so it makes sense. The fix is one click away.
Click any error, failing test or code issue to see what it is, why it matters and how to fix it. Then one button copies the exact prompt your AI needs to fix the root cause.
E-12Live37 times12 users
new row violates row-level security policy for table "orders"
- In plain language
- The database blocked the write: the table’s access policy doesn’t let this user insert or update that row. Fix: add a policy that lets each user save their own orders.
- Where it broke
- src/routes/checkout.tsx:84
- Where it came from
- Yesterday’s session, from the prompt “build the payment screen.”
Say it once. Every chat follows it. Each rule applies to your whole team.
Just say it: “from now on, never use axios.” Faundr recognizes it as a rule and saves it. From then on, the AI gets it automatically for everyone on the project, and nobody has to remember.
You set
- Rules always apply. “Payments only go through Stripe.”
- Decisions what you’ve already chosen, and why. The AI sticks with them until you ask for something else.
- Folder rules only kick in when the agent is about to edit a file there.
Rule says “never”? Faundr flags any edit that would break it, and the quality check points out where your code already does.
● Editing src/payments/checkout.ts
[Faundr] Team rules for src/payments/checkout.ts (they apply to this file; follow them):
- Payments only go through Stripe. Never store cards in the database.
- Amounts always in cents.
● Got it. Sending the amount to Stripe in cents.
The right context, at the right moment. You never have to repeat yourself.
The plugin tracks each moment of the session and feeds your AI only what matters right then. You never paste context by hand.
A session starts
Where you left off
A recap of your last session, the handoff note and your team’s rules.
You ask for something
Rules for the request
Relevant rules and decisions. New work gets planned before it’s built.
Before an edit
Rules for that file
Only the ones that apply, so the context stays lean.
Subagents
Same rules
Any subagent your AI spins up gets your team’s rules too.
When work wraps up
A handoff note is saved
What got done and what’s left, so the next session picks up right where this one ended.
Pick what shows under your prompt. Try it below.
See how much context you’ve used, what’s left in your 5-hour and weekly limits, and which step the work is on. Toggle items from the dashboard. It refreshes after every reply and uses zero tokens.
›
Faundr · Little Shop | Free shipping 2/5 | now: Shipping tests | map up to date | 1 concern | ctx ███░░ 42% | 5h █████ 84% resets 18:40 | wk ████░ 61%
From Faundr
- ProjectThe project connected to this folder.
- Current featureWhat’s being built and how many steps are done.
- Current stepThe step in progress, or what’s next.
- Code mapWhether the map is current or updating.
- ConcernsNumber of open concerns.
Plan usage
- ContextHow much of the context window is used. When it fills up, the agent compacts the chat and loses detail.
- 5-hour windowHow much of your 5-hour limit you’ve used (Pro and Max) and when it resets.
- WeekHow much of your weekly limit you’ve used (Pro and Max) and when it resets.
- Session costWhat this session would cost at API rates. Pro and Max plans are billed separately.
See how your app works, piece by piece. Click any piece to learn more.
Ask “draw the project’s architecture” or “draw how this will work.” Your AI maps out the parts, who handles each one and how data flows between them. Click a piece for a short explanation, where it lives in the code and whether it’s built or still planned.
Dashed outline = planned, no code yet.
Create session
Looks up prices in the database (never trusts what the browser sends), calculates the total in cents and opens a Stripe session.
- Comes from
- Payment screen
- Goes to
- Stripe · Orders
- Where it lives in the code
- src/server/payments.ts
Nothing happens behind your back. Every change can be undone.
AI agents move fast. The plugin makes sure you can follow along, verify and undo.
- Checkpoints
Faundr saves a checkpoint before the first change in every session, then every 30 minutes. Something broke? Roll back, and preview exactly what will change first.
- Risky commands flagged
Destructive commands, changes to your production database, packages with made-up names: the agent stops and asks first, with a checkpoint already saved.
- “Done” means done
Every feature has a “Done when” checklist. The AI checks it item by item before calling anything finished.
- Where did we leave off?
Everything still open: tasks, errors, security issues and the handoff note from your last session.
Your coding agent is powerful. With Faundr, you’re in control.
On its own, your agent assumes you know what to ask and can follow what it tells you. With the plugin, every step is explained and logged.
FAQ
Do I need to know how to code?
No. Commands come ready to copy, explanations are in plain language and you can ask for anything in your own words.
How do I install the plugin?
In Claude Code, paste /plugin marketplace add faundr-dev/faundr-plugin, then /plugin install faundr@faundr. In Codex, run codex plugin marketplace add faundr-dev/faundr-plugin and codex plugin add faundr@faundr in your terminal, then trust Faundr’s hooks with /hooks. The dashboard walks you through it when you create your project.
Does it use more tokens?
A little. Faundr picks the rules for each request and file locally, without AI, and only sends the ones that apply. The status line uses zero tokens.
Does it work for teams?
Yes. Rules, decisions and progress live in the project dashboard and sync to everyone’s AI agent.
Can I turn off the automations?
Yes, step by step, in the dashboard under Workflows → From Faundr. You can toggle each status line item too.