purpal is your product's decision memory. It pools every customer signal into themes, ties each call to its evidence, and flags when a problem you shipped a fix for comes back. Ask it here, or from Claude and your other AI tools.
A real run, start to finish
purpalConnects to where your signals live
AI coding agents are fast. But they still need someone to tell them what to build.
Customer complaints pile up in Intercom. Feature requests live in spreadsheets. What customers said on calls never makes it into the ticket. And six months later, nobody can say why the team chose what it chose.
purpal keeps the evidence. You make the calls.
A memory for every product decision.
purpal’s agents keep your evidence current around the clock, so every call starts from what customers said and what happened last time.
Always on·Pool every signal- 1
Pool every signal
Conversations, calls, reviews and files from every source, pooled into one set of themes and ranked by volume, urgency and growth.
conversations→themes
- 2
Weigh the evidence
Who is asking, how fast it is growing, what your analytics show and what the market is doing. Ask follow-ups in plain language.
themes→evidence
- 3
Record the decision
Every claim cited to a customer or flagged as an assumption. When you say go, the spec and tickets follow.
evidence→a cited decision
- 4
Watch the outcome
After you ship, purpal keeps listening. If the complaints come back, the theme reopens and lands on your Home.
a release→an outcome
From scattered feedback to a decision you can defend.
What customers said across every tool becomes cited evidence for the call, with the spec ready when you decide to build.
“...they need faster search but also mentioned wanting filters, and then the PM from Enterprise jumped in about sorting...”
Who owns this? No action item created.
@channel can we also add sorting? The PM from Enterprise asked for it yesterday. Also @design is the search bar moving?
3 different threads. No resolution.
PRIORITY: Fix search!! But also Jane said filters are more important?
Problem
Users abandon search due to lack of filtering, causing 34% drop-off and 67% bounce on results pages.
Requirements
- Faceted search with category filters
- Real-time suggestions (<200ms)
- Sort by relevance, date, popularity
User Story
“As a power user, I want to filter results by category so I can find relevant items faster.”
Acceptance Criteria
- Results update within 200ms
- Filters persist across sessions
- Mobile responsive, 44px+ touch targets
AI coding agents solved “how to build.”
purpal solves “what to build.”
How to build
AI coding tools ship features in hours. The implementation is no longer the bottleneck.
What to build
Which problem, backed by which evidence, worth which trade-off. No coding agent answers that.
When the decision is wrong, you don’t just waste time. You ship the wrong thing faster.
Writing the spec is the easy part now. Knowing whether it should exist is the hard part. purpal keeps the evidence, and every past decision, one question away.
One memory, for your team and your AI tools.
The same evidence, readable by your team and queryable by the AI tools they already use. No copying context between tabs.
- See who is asking, how fast it is growing and what analytics show before you commit.
- Every claim is cited to a customer or flagged as an assumption.
- Decisions, terms and constraints are remembered, so no one relitigates last quarter.
Acceptance Criteria
- Results update within 200ms
- Filters persist across sessions
- Mobile responsive, 44px+ touch targets
Evidence
“Search is useless without filters.” signal · intercom
34% drop-off on results pages assumption · unverified
How it works
From customer pain to a decision, and back again.
- 1
Connect your signals
Connect Intercom, upload a CSV, or pipe in signals from any source. purpal pools them into themes and surfaces what is growing and urgent.
and more - 2
Decide with evidence
Open a theme to see the customers, trend and analytics behind it, and ask follow-ups in chat. When you commit, the spec is drafted with every claim cited.
Smart Search EnhancementCitedApprove - 3
Ship and watch
Push to Linear, or hand off to coding agents. purpal keeps listening after release and reopens the theme if the complaints return.
ENG-412Faceted search filtersENG-413Sort by relevanceShipped · watching for regressions
of claims cited or flagged
listening after you ship
tools your AI agents can call over MCP
Built for PMs who decide, not PMs who assemble.
Good fit
- You spend more time assembling evidence than making decisions.
- Your team uses AI coding tools and needs to know what is worth building.
- You ship weekly and can’t afford building the wrong thing.
- You want customer evidence behind every call, not gut feel.
Not a fit
- You prefer to figure requirements out in code.
- Your team doesn’t have a PM or product function.
- You only want AI to write specs faster.
Better to know now than three weeks in.
Know what to build. And why.
purpal is live and onboarding teams one at a time. Join the waitlist to give your team a product memory that every decision, and every AI tool, can draw on.
Free to start · No credit card · Onboarding teams one at a time