01 · Goals
Mission, objectives, and key results — mapped in a hierarchy where every task connects back to the outcome it serves.
→ live in graph
The engine
Built to understand.
Designed to compound.
Every decision Humfrid makes is grounded in your business — and gets sharper with every action.
Shared context
Knows your team.
Everything your team learns, decides, and reasons through — kept, connected, and ready the next time it matters.
01
Extracts what matters.
Reads across your analytics, code, interviews, and competitor moves, and pulls out the signal worth remembering.
02
Always aligned.
Every memory is checked against last week's research, standing decisions, and every teammate's plan. In seconds. Not in next month's retro.
03
Every decision, at a glance.
Captured as you make it — and connected to the data and learnings that led to it. The why never gets lost in a doc nobody can find.
Everything, in one graph.
Goals, customers, code, principles, decisions, and product — structured, linked, and understood by Humfrid.
02 · Customers
Every customer, every conversation, everything they've said — connected. Not a CRM field. The actual signal, attached to the person who gave it.
→ live in graph
03 · Code
Your codebase as a graph, not a directory. Every function mapped to the decision that shaped it. Every change scored for blast radius before it lands.
→ live in graph
04 · Principles
The rules your team operates by, learned from how you actually work. New decisions get measured against them automatically.
→ live in graph
05 · Decisions
The calls your team has made, and why. Context and tradeoffs logged and carried forward. No re-litigating settled questions.
→ live in graph
06 · Product
Every feature, what users have said about it, and the decisions that shaped it. The full history of why the product looks the way it does.
→ live in graph
↳ Sources feeding the graph
Customer calls Voice notes Strategy docs PostHog Google Analytics GitHub more integrations coming soon
Memory
Memory that means something
Most tools store. Humfrid understands.
$ |Capture
Every format you use
Text, voice, PDF, DOCX, PPTX, Excel, HTML, CSV, email. If you have it, we parse it.
Audio with speaker separation
Upload audio files and get automatic transcription that knows who said what. Voice-to-context in one step.
Connected sources
Slack threads, Jira tickets, Google Drive — with full attribution preserved. Changes sync automatically.
Extract
4-phase pipeline
Raw text → atomic facts → named entities → deep extraction of what’s been decided, assumed, and learned.
Auto-classification
Every statement automatically categorized — data point, lesson learned, decision, or guiding principle. With category, author, title, and tags detected automatically.
Automatic entity resolution
Mentions of people and companies resolve to the right entity automatically — even across nicknames, typos, or partial names.
Duplicate detection
Semantic deduplication before anything enters the graph. Upload the same content twice? Caught automatically.
Structure
Product model
Define your actual product: features, capabilities, user segments, accounts. Every data point lands in the right context automatically.
Source and certainty
Track who said what — and how confident they were. Was it a firm decision, a working assumption, or someone’s gut feeling? Every piece of context carries its source and certainty level.
Derivation chains
Every decision links back to the research and signals that shaped it. You can always see why a call was made — not just what it was.
Connections
Context isn’t stored in isolation. Each piece connects to others — building on, supporting, contradicting, or replacing what came before. The relationships are explicit, not implied.
Cross-session persistence
What you learned yesterday is still there tomorrow — across every agent. Just pick up where you left off and watch your agents get smarter every day.
Semantic search & retrieval
Entity-aware semantic search across everything your team knows. Ask about a person, project, or decision and get precise answers with sources.
Under the hood
Built to be questioned.
Discrete data entities with derivation chains and provenance. Every decision, principle, signal, and goal traces back to the source it came from — and forward to everything it influenced.
Derivation chains
Every atom shows where it came from and what it's derived from. A decision traces to the meeting it was made in. A principle traces to the five decisions that revealed it.
Provenance
Every claim is backed by its source — the exact sentence in the transcript, the message in the thread, the line in the doc. Click any atom, see the receipt.
Versioned
Atoms evolve as your team does. A decision that gets revisited isn't overwritten — it's versioned, with the reasoning for the change captured alongside it.
DECISION Jan 12 · Arthur
Go with per-seat pricing over usage-based model.
Reason: usage-based created unpredictability for finance teams. Per-seat easier to budget.
Derivation chain
DATA 5 customer interviews flagged pricing confusion
↓ derives from
LEARNING Finance teams budget by seat, not by usage
↓ informed
DECISION Per-seat pricing ← this atom
↓ shaped
PRINCIPLE Pricing must be predictable for buyers
Source
📄 pricing-strategy-review.pdf · Jan 12
"...after reviewing the interviews, we're going with per-seat. Usage-based was confusing finance teams and stalling deals."
Version history
v2 · current Jan 12
Per-seat pricing.
v1 · superseded Dec 3
Usage-based pricing. Revisited after customer interviews showed budget friction.
Contradictions surface at authorship.
Not after launch.
When a new decision contradicts an existing principle or data point, agents flag it immediately — while it still costs nothing to fix.
Decision
“Build advanced dashboard analytics — prioritized for Q2.”
conflicts with
Data
“Analytics never mentioned in 7 of 10 recent customer calls. Search was #1.”
From outcome to task
Thinks like your best
product manager.
Keeps track of your objectives and metrics, maps the opportunity space, identifies solutions, and suggests how to implement them.
Proactive intelligence
Just know.
No dashboards to check. No meeting summary you forgot to read. When something matters, Humfrid tells you — wherever you already are.
Know before your CEO does.
A customer complained. A metric moved. A pattern emerged. You hear it first, with the evidence.
Know before it's too late.
Another team made a decision that touches your roadmap? Humfrid flags the conflict the moment it happens — not in next month's retro.
Know without asking.
A morning briefing in your channel before you've made coffee. Updates when something matters. Silence when nothing does.
Works where you do
Right where your team already works.
Humfrid lives in Slack and plugs into the tools your work already runs on — so there's no new place to check.
# product
HumfridApp8:30 AM
Quick follow-up on the activation goal — the onboarding tour you shipped two weeks ago is paying off. Activation's up 3pp since launch (41% now), so you're nearly halfway to the 45% target and a little ahead of pace.
🙌 3 🎉 2
DN
Daniel9:02 AM
@Humfrid nice! should we double down and add a second tour step?
HumfridApp9:03 AM
Worth thinking about both sides. I just checked our funnel analytics — people who finish the tour activate about 19% higher (last 1,200 sign-ups), so there's clearly value in it. But 22% drop off on step 1, and a second step would probably make that worse. I'd tighten step 1 first before adding to it. Should I add it as an opportunity and write a brief?
DN
Daniel9:04 AM
@Humfrid yes let's do it!
PR
Priya9:05 AM
nice work team! 🎉
Linear - Turns objectives into issues and cycles
- Keeps the roadmap matching reality
GitHub - Reads your codebase as a graph
- Opens pull requests
- Ships straight to your app
Google Analytics - Watches your targets
- Reports what moves
- Spots the drop-offs
Slack - Briefings, blockers, and approvals
- Right in the thread you're already in
Figma - Reads your designs and flows
- Keeps what ships true to them
Google Drive - Grounds decisions in your docs, specs, and decks
- Stays current as your files change
Plus 
For the engineers on your team
Your context,
in their editor.
Through MCP, every coding agent your team runs pulls its brief straight from Humfrid — the objective, the decisions, the specs. So what they build is what you meant.
01
Connect via MCP.
Humfrid is native to the protocol your agents already speak. Claude Code, Cursor, Codex — plug in, and your context shows up.
02
Assign tasks with full context.
Assign an agent and it gets handed the goals, metrics, tasks, decisions, principles, and customer signals it needs — automatically attached. No more prompting.
03
Agents store their decisions, too.
Every session leaves a trace. Everything the agent learned and decided along the way — all captured, all queryable by the next session and by every other agent.
Works with 
>what should I work on next?
●humfrid · recall("objective · cut first-week churn")
⎿ 3 bets, ranked by expected impact:
1 · nudge a 2nd data source by day 3
2 · fix the step-3 onboarding drop-off
3 · re-engage 1,200 dormant trials
>1 — why this one?
●humfrid · search("2nd-source · churn evidence")
⎿ 71% of churned accounts never connected a 2nd source
not in any dashboard — found across calls + usage logs
Here's the evidence, the open question, and an experiment to settle it. Draft the spec?▋
Signals in. Impact out.
Finally, the hardest job
in tech gets easier.
Humfrid helps you move from signals to shipped to impact, with confidence and ease.
01 · Signals
Every click, user interview, and online review — read continuously, the moment it happens. Nothing worth knowing slips past.
02 · Opportunities
Scattered signal becomes a tree of opportunities — each one traceable to the evidence behind it. You see the reasoning, not a black-box score.
03 · Ship
The work to capture it gets planned, delegated, and built — by your team and your agents, all from the same shared context.
04 · Measure
How success will be measured is defined before anything ships. Then the metric is watched automatically — no dashboard to remember to open.
05 · Learn
What worked feeds the next move. What didn't, too. The loop comes back around sharper every time — and never forgets why.
Outcomes
Automatically improve your product metrics.
Finally, the hardest job in tech gets easier. Humfrid helps you move from signals to shipped to impact, with confidence and ease.
Pressure-test your assumptions
Check what you're about to ship against what the numbers show, your users actually say, and what you're aiming to achieve, so you make the best call.
Identifies opportunities
Humfrid watches your metrics and conversations for untapped upside — a segment converting better, a feature driving retention — and shows you where the next win is.
Automatic follow-up on every release
Humfrid follows up on every release — checks the outcome against your hypothesis, tells you what actually moved the metric, and flags anything drifting before it's a problem.
Learnings that compound
Every release, call, and decision feeds back in. Humfrid captures what your team learns and links it to what you already knew, so your product knowledge compounds instead of starting over.
Stay aligned without the meetings
Humfrid automatically checks for contradictions in data, decisions and reasoning between teams, and alerts you before they become a problem.
Keep track of all decisions
Ask it anything about your product, your team, or what you decided. Humfrid keeps track of everything and answers immediately, with context everyone can see.
The AI-native product org
Your engineering team runs on AI now.
Your product team should too.
When your engineers ship faster, it becomes more important than ever to stay on top of goals, customers and data. Humfrid makes sure you do.
Without Humfrid
With Humfrid
Product managers spend most of their time chasing information instead of deciding.
Data, customer learnings, previous decisions, product principles, goals and plans - searchable and queryable from one place.
Product managers create plans, PRDs, and tickets with LLMs — with mediocre results.
Product managers create impactful plans, PRDs, and tickets grounded in your shared context.
Product managers spend endless hours in alignment meetings.
Humfrid proactively alerts you when your decisions don't match the data, or when another team is making conflicting plans - so you can course-correct immediately.
Engineers ship the wrong thing faster than ever.
Claude Code and Cursor read your product context through MCP before they write a line of code.
You only discover in QA that engineers built the wrong thing.
Every PR is reviewed to make sure it adheres to what customers want, product decisions, and team principles.
You can't tell whether the AI-built feature actually worked.
Every shipped feature automatically measured against the outcome it was built to impact.
You run experiments to learn - but struggle to systematically apply the learnings to your next bet.
Every learning flows back into your shared context, so every team and agent builds on it. Insight compounds with each bet.
Without Humfrid
Product managers spend most of their time chasing information instead of deciding.
With Humfrid
Data, customer learnings, previous decisions, product principles, goals and plans — searchable and queryable from one place.
Without Humfrid
Product managers create plans, PRDs, and tickets with LLMs — with mediocre results.
With Humfrid
Impactful plans, PRDs, and tickets grounded in your shared context.
Without Humfrid
Product managers spend endless hours in alignment meetings.
With Humfrid
Humfrid alerts you when your decisions don't match the data, or when another team is making conflicting plans — so you can course-correct immediately.
Without Humfrid
Engineers ship the wrong thing faster than ever.
With Humfrid
Claude Code and Cursor read your product context through MCP before they write a line of code.
Without Humfrid
You only discover in QA that engineers built the wrong thing.
With Humfrid
Every PR is reviewed to make sure it adheres to what customers want, product decisions, and team principles.
Without Humfrid
You can't tell whether the AI-built feature actually worked.
With Humfrid
Every shipped feature is automatically measured against the outcome it was built to impact.
Without Humfrid
You run experiments to learn, but struggle to systematically apply the learnings to your next bet.
With Humfrid
Every learning flows back into your shared context, so every team and agent builds on it. Insight compounds with each bet.
Conflict detection
When decisions don't add up, know instantly.
Standups, planning docs, OKR reviews — meetings that exist because context doesn't travel on its own. Humfrid makes it travel. Misalignment caught — explicit and implicit — at authorship, not at launch.
DECISION
"Build advanced dashboard analytics — prioritised for Q2."
⚡ conflict
DATA
"Analytics never mentioned in 7 of 10 recent customer calls. Search was #1."
Build what your users actually want.
Humfrid flags when a new decision contradicts what customers have told you. You hear it in the planning doc, not six weeks after shipping.
Stay true to your principles.
Your team agreed that every new feature ships behind a flag until validated. A new PRD quietly skips the flag. Humfrid catches it immediately — not after the rollback.
Goodbye silent misalignment.
Two teams, two decisions, two directions. Humfrid catches the drift the moment the second decision is written — while it still costs nothing to reconcile.
Code intelligence
Run coding agents at full power.
Humfrid indexes your code alongside your decisions, principles, and customer signals — so the "why" and the "what" finally are connected.
20x+ faster than Grep.
Every function, every caller, every downstream dependency — indexed and queryable in milliseconds. Your agents stop reading 20 files blind and start with the answer.
Blast radius before you touch a line.
Risk level on every change. All surfaced before the edit, not after the regression.
Decisions linked to the code they shaped.
The "why" lives next to the "what." When an agent proposes raw SQL, it sees the team already decided to use the ORM. Architectural drift, caught at authorship.
Multi-agent coordination, live.
Two agents editing the same file used to be a merge conflict waiting three hours to happen. With Humfrid, they see each other the moment they both reach for the same symbol.
Every product team's favorite agent.
Humfrid is the friendly product agent that helps your team spend time on what matters.
Free to start · No credit card