Your AI remembers conversations.It still doesn't understand users.
Percipio continuously builds an evolving understanding of every user, so your AI becomes more personal, contextual, and useful over time.
Built for AI products. Integrated in an afternoon.
usr_92aThe recurring problem
Every AI team eventually builds the same thing.
Models can reason over a prompt. They cannot maintain a coherent understanding of a person across months of interaction.
Conversation memory
Store transcripts, summarize threads, retrieve fragments.
User profiles
Turn scattered facts into something the product can use.
Personalization
Decide what matters now and what should persist.
Context engineering
Fit the right understanding into every model call.
Percipio replaces thousands of lines of custom infrastructure with a few API calls.
How it works
Conversation becomes understanding.
Evidence in. Structured state out. Better context on the next model call.
Send conversations
Send ordered user and assistant messages after each exchange.
POST /v1/conversationsUpdate the Human Model
Percipio extracts evidence and deterministically merges typed state.
goals: +1, preferences: updated}Retrieve optimized context
Fetch concise, structured understanding before calling your LLM.
GET /v1/users/{id}/contextFrom signal to state
A model of the person—not a pile of messages.
Click a message to see where explicit user evidence lands in the Human Model.
{
"goals": [{
"title": "Prepare for YC",
"confidence": 0.97
}],
"projects": [{
"name": "Seed fundraising"
}],
"communication_style": {
"preference": "concise"
}
}goals[0]/v1/users/usr_92a/context200 · 84msPlatform primitives
The infrastructure understanding requires.
Persistent user understanding
A living model that compounds across every interaction—not another transcript store.
Structured Human Models
Typed goals, projects, preferences, relationships, interests, and memories.
Context retrieval
Prompt-ready context ordered by category, confidence, and recency.
State transitions
Every conversation becomes a deterministic, traceable update to user state.
Version history
Immutable snapshots and events make every change inspectable and recoverable.
Confidence scoring
Uncertainty stays explicit. Repeated evidence strengthens what the model knows.
Developer-first API
One ingestion endpoint. One context endpoint. Clear schemas and stable semantics.
Privacy-first architecture
Organization isolation, protected-trait filtering, deletion APIs, and user export.
Launch-ready controls
Rate limits, idempotency, status checks, and metadata-only request logs.
Data rights built in
Export and delete end-user data through simple API endpoints from day one.
Trust posture
No training on customer data, 90-day raw-message purge, and transparent subprocessors.
Developer experience
Useful context. One request away.
Percipio sits between your conversation store and your model. Keep your stack. Improve what it knows.
- Synchronous state transitions
- Deterministic context snapshots
- Organization-scoped API keys
- Export, deletion, and rate-limit controls
const context = await fetch(
`${process.env.PERCIPIO_URL}/v1/users/${userId}/context`,
{
headers: {
Authorization: `Bearer ${process.env.PERCIPIO_API_KEY}`,
},
},
).then((response) => response.json());
const response = await openai.responses.create({
model: "gpt-5.4",
instructions: context.text,
input: userMessage,
});Memory is not understanding
Retrieval finds what was said. Percipio tracks what changed.
| Capability | Traditional memory | Percipio |
|---|---|---|
| Stores conversations | ✓ | ✓ |
| Understands goals | — | ✓ |
| Tracks projects | — | ✓ |
| Updates preferences | — | ✓ |
| Maintains relationships | — | ✓ |
| Adapts over time | — | ✓ |
| Structured retrieval | — | ✓ |
The long view
The missing layer of AI.
Language models transformed how machines understand language.Percipio helps them understand people.