Concepts and vocabulary

The words you'll see in the dashboard and in the commands, explained one by one.

Organization, project and session

Organization

The group of people and projects. Every project belongs to an organization, and every member of that organization can see all of its projects. See Organization, team and account.

Project

A product you're building: an app, a site, an API. In the dashboard, each project has its own sections (Overview, Features, Security…).

Linked folder

The code folder on your computer that was connected to a project with /faundr:link. It contains a .faundr.json file. Faundr only acts in linked folders.

Session

One conversation with your AI agent in a linked folder, from the moment you open it until you close it. Each session shows up in Activity → Session log with the requests, changed files, commits and completed tasks. Requests and replies stay with the person who ran the session; the team sees the summary.

Plugin

The part of Faundr installed inside your AI agent (Claude Code or Codex). It watches sessions, hands over context, runs the checks and adds the /faundr:… commands ($faundr:… in Codex).

Content language

The language the agent uses for everything it records in Faundr for that project (Overview, Stack, features, decisions, findings, handoff notes), whatever language each person chats in. You set it under Settings → General. Anyone reading the dashboard in another language can use the Translate button, which runs Chrome's built-in translator on your computer and costs no tokens.

The work

Feature

Something the product will do, like "Login screen" or "Card payments". It has a task checklist, done-when criteria and the decisions tied to it.

Current feature

The feature you're working on in that folder. New tasks go into it, and the agent receives it at the start of each session. To switch: faundr focus "<name>".

Feature owner

Whoever created or started the feature. Its tasks follow the owner: if someone else's agent tries to change them, it gets a warning and only goes ahead if that person insists, after checking with you. To change owners: faundr own and faundr disown.

Stalled

A feature in progress that nobody has touched for more than 14 days. /faundr:tidy-board checks stalled features against your commits and code and proposes finishing, discarding or keeping each one.

Discarded

A feature that won't happen. It leaves the tabs (it stays under All) and the agent's context, and comes back with Restore.

Milestone

A phase of a large piece of work. Each milestone is a feature that "comes after" another one. Until work on it starts, a milestone shows as Planned under Next milestones.

Task

One step on a feature's checklist. It can be to do, in progress or done.

Uncommitted

A task marked done while there were changes that hadn't gone into git. The mark clears on its own at the next git commit. It reminds you that the work isn't saved in history yet.

Done-when criteria

The list of what has to be true for the feature to be finished, like "the customer pays by card and gets the email". When the checklist is complete, the agent checks each item and marks it met or not met.

Planning

What the agent does before building something new (/faundr:planning). It asks a few questions about what only you can decide, sets a "done when" you can verify and gives you a summary to approve. Your choices become decisions.

Memory

Rule

Something the agent must always follow, like "every server route checks project access". It goes to the agent at the start of each session and to its subagents.

Decision

A technical choice that's already been made, with the reason, like "payments through Stripe because it supports local payment methods". It reaches the agent when it matters: at the start of the session (the newest ones and the ones for the current feature), when a request touches the same topic, or before the agent edits the files tied to it. That way the agent won't re-propose something the team already settled.

File-scoped rule or decision

A rule or decision that applies only to certain files or folders. It reaches the agent right before it edits one of those files instead of at the start of every session.

Forbidden pattern

A piece of text a rule says the code must not contain (for example, console.log). Faundr warns the agent on every edit that introduces it, and the quality check points out where it already exists.

Proposal

A rule or decision that hasn't been confirmed yet. It doesn't reach the agent until someone clicks Confirm.

Obsolete

A rule or decision that no longer applies. It's kept on record but no longer sent to the agent.

Concern

Something to review later that's neither a rule nor a task, like "login may get slow with lots of users". It gets a P-n number and stays highlighted until someone resolves it.

Note to agents

A message written in the dashboard (Activity → Now, type Note to agents) that reaches the agent along with the next request from whoever is in a session. Example: "don't touch the payment module".

Skills and workflows

Skill

A playbook the agent follows for a type of task. The /faundr:… commands are Faundr's skills. Your project, each person and other plugins can have their own.

Project skill

A skill you already use that /faundr:skills has read. Faundr stores its name, description, source, when to use and topics, and reminds the agent about it when a request touches one of those topics. It shows up under Workflows → Skills. See Workflows.

Connected skill

A project skill whose job is to find problems: security, tests, code quality or errors. It gets the Connected label. When it runs, the agent sends Faundr what it found and fixed at the end of the response, and the problems become S-n, Q-n, T-n or E-n tagged via /name. Each skill only closes its own findings.

Workflow

The playbook your team wants the agent to follow for a type of request ("whenever it's about design, do this, then that"). It starts as a draft and only takes effect once someone approves it in the dashboard, under Workflows. It gets a number, W-n.

Continuity

Handoff note

One to three sentences about what's half done, what the next step is and anything the next person should watch out for. The agent writes it at the end of long sessions. It appears in Activity → Where you left off, on the project card and at the start of the next session.

Handoff pack

A complete document a developer can use to take over the project: what it is, where it stopped, where it runs, how to run it, rules, decisions, risks and the code map. It lives in Activity → Reports.

Restore point

A snapshot of the project's files, stored in git, that you can go back to if something goes wrong. Faundr saves one automatically before changes. It doesn't include .env, node_modules or the database. See Disaster protection.

What Faundr finds

Everything Faundr finds becomes a finding with a letter and a number. The number never changes, so you can refer to a finding anywhere: "fix S-3".

Findings that come from a connected skill carry that skill's via /name tag.

Findings have a status:

• Open: needs attention.
• Fixed or resolved: the problem is gone. In most cases Faundr only closes a finding after re-running the check and confirming the problem no longer shows up.
• Ignored: someone decided it isn't a problem, or that it can wait, and gave a reason. Some come back on their own after a set period.
• Archived: doesn't apply to this project. It won't come back.
• Regressed: it was fixed and showed up again.

Severity

How much the finding matters: critical, high, medium or low. In the dashboard, severity shows through the label's intensity. The more severe, the brighter.

Suppressed

A security finding Faundr considered noise, like a sample key in a test file. It's hidden, and you can see it with Show suppressed.

How Faundr checks

Check without AI

The automatic analysis that runs on your computer using fixed rules. It doesn't use the agent and costs no tokens. It runs on its own at the start of sessions and after responses that changed files.

AI review

The analysis done by the agent itself when you ask for it (/faundr:security, /faundr:quality, /faundr:design). It reads the code's logic and finds what fixed rules can't see. It uses tokens from your AI plan.

Code map (or graph)

An index of the project built by Faundr: the files, the functions and what uses what. It helps the agent find things while reading less, and it helps you see how the project is organized. See Code map.

design.md

The project's design guide: colors, fonts, components and what to avoid. Faundr checks your screens against it.

Overview

The project summary written by the agent: what it is, who it's for, how it works, its parts and where you are.

Stack

Where each part of the project runs and on what: hosting, database, login, languages, external services and variables.

Tokens

The unit an AI model uses to measure (and bill) the text it reads and writes. Insights shows how many were used and how much of that came from Faundr.