Beta · Early Access Open

Make your
Product f()unction

The AI-powered Product Operating System. Every capability is an executable function backed by persistent team memory, evidence chains, and human oversight. Now in private beta.

✓ You're in. We'll send your beta invite soon.
// the product core

How Product f() works.

An AI-powered Product Operating System for product managers, founders, VPs, and builders. Every AI call starts where your team left off — aware of your product, data, and past decisions.

Connectors

Your stack, normalized

GA4, Stripe, Google Sheets, CSV, Mixpanel, HubSpot, Postgres — with source lineage, sync timestamps, and connector health.

See how it flows → The Brain

Persistent team memory

A privacy-controlled memory layer beneath every f() module. Nine categories, progressively injected into every call.

Explore the Brain → f() functions

Named, executable AI

Every capability is an executable skill — with its own intake questions, connector requirements, and run history.

What is f() → How it works

Context → confidence → HITL

Brain assembles context, an agent runs, outputs are confidence-scored, and humans stay in control of every Class-2 action.

See it in action →
// see it in action

A working product.

Live in private beta. Swipe through the core surfaces.

AI Assistant
AI Assistant with multi-model support

Chat with your product data. The AI knows your context from the Brain.

f() Functions
Executable product functions

Every capability is a named, executable function — one click away.

The Brain
Knowledge graph with 852 nodes

Persistent team memory — product identity, research, decisions, strategy.

Get Started
Quick product onboarding

Connect your stack, set your product identity, the Brain learns immediately.

// what is f()

A product OS built
around functions.

In software, a function takes an input, applies intelligence, and returns a reliable output. Product f() applies this idea to your entire product workflow. Each capability is a named, executable function deterministic in its constraints, intelligent in its reasoning, and always reviewable by a human.

It's not a suite of AI tools. It's a coordinated operating system where every module shares the same memory, the same evidence chain, and the same respect for human judgment.

⛓

Evidence before action

Every output, recommendation, and action is traceable to a real evidence source: KPIs, research, tickets, decisions.

👁

Human judgment at meaningful decision points

High-stakes actions always surface for review. The system never becomes a black box that acts without your approval.

📡

Continuous background intelligence

Not session-only AI. The OS monitors signals, prepares drafts, detects risks while you're working on other things.

🔍

Comprehension-first outputs

You can always see what the AI knew, what it reasoned, and how confident it is. Trust is earned, not assumed.

// the f() universe

Every function is an executable skill.

A live look at the active f() catalogue — each function carries its own skill definition, intake schema, connector requirements, and run history. Swipe through a few; click any card to inspect its skill.

// f(brain)

The only AI that
never forgets your team.

The Brain is a secure, persistent memory layer beneath every f() module. It accumulates context from every task, query, decision, and insight your team generates, then injects that context intelligently into every future AI operation.

Without the Brain, every AI call starts from zero. With the Brain, every AI call starts from where your team left off, aware of your product, your ICP, your past decisions, your KPI baselines, and your preferred ways of working.

Company Identity
Level 0 · always injected. Mission, values, business model.
Product Identity
Name, ICP, stage, positioning, differentiators
Team Identity
Members, roles, ownership domains, rituals
Strategy Memory
OKRs, roadmap commitments, competitive positioning
Research Memory
Validated personas, pain points, recurring insights
Data Memory
KPI baselines, anomaly patterns, what "good" looks like
Decision Log
Past decisions with rationale, owners, and outcomes
Learned Preferences
Frameworks, tone, PRD format, inferred from usage
Active Context
Current sprint, top priorities, blockers, experiments

// 852 nodes of team memory

Brain knowledge graph — 852 nodes
Product Data Research Decisions

// progressive context injection

Level 0: Company + Product identity, always on (<800 tokens).
Level 1: Task-relevant strategy, decisions, KPIs.
Level 2: Deep semantic recall via pgvector similarity.
Level 3: Full history — only when the user explicitly asks.

// human-in-the-loop

AI that earns trust.
Humans that stay in control.

Product f() is the first product OS designed around the principle that AI velocity must never exceed human verification capacity. Every agent action is classified, and high-stakes actions always surface for review before execution.

⚙️

Two-Class Action Model

Class 1 (Read & Generate) runs freely and is fully logged. Class 2 (Write & External) sending Slack messages, creating tickets, sending emails, always enters the HITL queue before execution.

🔎

Full Evidence Transparency

Every AI-generated output shows exactly which Brain entries were used, which model produced it, and the confidence score. A collapsible "Brain context used" panel is shown by default on every output.

📋

Configurable Review Modes

Set your review posture per task: Ambient (complete & save), Final Review (review then save), or Ongoing Review (checkpoint-by-checkpoint). You decide the oversight level.

🤖

Auto-Approve Rules

Create rules for trusted, low-stakes actions, always auto-approve Monday pulse to #product. Hardcoded restrictions prevent auto-approval of external emails, ticket changes, or irreversible operations.

📝

Brain Writes Are Staged

Agent outputs never write directly to your team's memory. They write to a staging area first. You see a summary of what the system wants to learn, then approve, discard, or selectively save entries.

🛡

Deterministic Guardrails

Before AI reasoning even runs, constraint validation scripts enforce structural requirements, correct owner assigned, evidence present, minimum data windows met. AI handles pattern recognition; scripts handle correctness.

// who it's for

Built for product teams
that operate at high stakes.

IC Product Manager

The PM who needs answers fast

  • Ask your data in plain language
  • Generate PRDs grounded in evidence
  • Run launch gate reviews with AI assistance
  • Stop re-entering context into every AI tool
Founder / CPO

The leader who needs portfolio clarity

  • Monitor product health across all signals
  • Detect risks before they become incidents
  • Audit every AI recommendation's evidence chain
  • Make decisions backed by your team's full history
VP Product / Director

The leader who needs cross-team visibility

  • Standardized operating rhythms across teams
  • Sprint health and release confidence at a glance
  • Decision traceability across quarters
  • Less status theater, more ambient intelligence
Growth PM / Marketing PM

The PM who needs true attribution

  • True CAC, not just last-touch estimates
  • Channel performance across all acquisition sources
  • Product and growth data in the same query
  • Research signals feeding directly into campaigns
// how it works

Connect your stack.
Let your OS learn. Stay in control.

Three simple steps to a product team that operates 10× within its capabilities.

01

Connect your stack

GA4, Stripe, Sheets, Linear, HubSpot, Postgres — normalized with source lineage and connector health.

02

Brain assembles context

Nine memory categories, four disclosure levels. The Brain injects only what this task needs.

03

Run your f() function

A named, executable AI behaviour runs via LangGraph — grounded in your live evidence chain.

04

Confidence scoring

Every output is scored. ≥0.85 auto-applies. 0.60–0.85 enters HITL review. <0.60 is discarded.

05

HITL + audit trail

Class-2 actions always surface for review. Every artifact is saved with its Brain context and reasoning.

Slack
GitHub
Stripe
Google Drive
Notion
Google Analytics
Linear
Jira
Slack
GitHub
Stripe
Google Drive
Notion
Google Analytics
Linear
Jira
f()
// beta access

Join the private beta.
Make your product f()unction.

Product f() is live and accepting beta users. Join now to get early access, shape the roadmap, and start building your team's persistent AI memory.

✓ Welcome to the beta. Check your inbox for next steps.
⛓ Evidence before action
🔍 Comprehension before automation
🔎 Progressive disclosure
👁 Human judgment at decision points