Introducing Agent Brain.

A private, company-owned knowledge base with grounded citations and durable semantic memory — the context agents reason with, and it compounds across every run.

naive upprovisioning
Postgres

postgres://agent@db.usenaive.com:5432

primary
Auth

JWT · sessions

100%
Storage

Object storage

100%
Realtime

live sync

88%
Edge fn

serverless

72%
Queue

SQS · workers

54%
Cron

scheduled

36%

$ naive up --plan

7 services1.2s elapsed

Paste and ship

Drop this into Cursor, Claude Code, or any agent. It reads the skill manifest and wires up Agent-Native Brain for you.

Agent prompt

Read https://usenaive.ai/skill.md and set up Naïve for Agent-Native Brain in my project.

Knowledge and memory agents can trust.

Ingest documents once, ask grounded questions with citations, and write durable facts that survive the run — Postgres-owned end to end, with provable deletion.

Semantic Memory

Facts and entities that compound.

Read docs

Knowledge Graph

Entities, relations, and timelines.

Read docs

Context agents can cite and remember.

Ingest once, answer with citations, and keep durable memory future runs inherit — one company-scoped knowledge base, private by default.

POST /v1/brain/query

Document RAG

Ingest text, URLs, or files once, then ask grounded questions and get answers with citations that map back to your own documents — a citation that can't be traced to your sources is dropped, so answers never leak another tenant's data.

Ingest anything
Text, URL, or file.
Grounded answers
Citations to your docs.
Sessions
Continue the thread.
Company-scoped
Private by default.
$ naive brain add --file ./handbook.pdf
$ naive brain query "What is our refund policy?"
POST /v1/brain/remember

Semantic Memory

Write durable facts, entities, and relationships as the workforce operates, then recall or reason over them on later runs — memory that compounds instead of resetting each session.

Remember
Facts from any run.
Recall
Pull them back later.
Think
Reason over memory.
Compounds
Survives the run.
$ naive brain remember "Acme prefers quarterly invoicing" --source crm
$ naive brain recall "Acme invoicing"
$ naive brain think "How should we bill Acme?"
POST /v1/brain/graph

Knowledge Graph & Timeline

Explore the entities and relationships Brain has learned and walk the timeline of what was true when — every claim carries provenance back to the run that produced it.

Entities
People, orgs, facts.
Relations
How they connect.
Timeline
What was true when.
Provenance
Traceable claims.
$ naive brain graph --entity "Acme Corp"
$ naive brain timeline --entity "Acme Corp"
Governance

Grounding & Provable Deletion

Destructive operations are approval-gated for agent callers, deletes return a provable DeletionReceipt, and agent-written memory can route through a proposal → review → promotion gateway so canonical company memory is never silently mutated.

Approval-gated
Sign-off to delete.
Provable deletion
Confirmed + verified.
Memory Gateway
Propose, review, promote.
Opt-in
Off until enabled.
$ naive brain forget episode ep_5533 --reason "superseded"   # 202 + pending approval

Build agents that remember and cite.

A private knowledge base with grounded answers, durable semantic memory, and provable deletion — one API.