Northcontext

A memory layer your AI app can plug into in 5 minutes. Multi-tenant, semantic, audited. Runs on Cloudflare.

no card. demo mode runs in your browser.

1.	Store conversations, get back facts — northcontext.dev/docs/extraction your app sends raw transcripts, we extract typed memories (preference / fact / goal / event / relationship) and embed them so semantic recall just works.
2.	One endpoint to inject relevant memory — northcontext.dev/docs/recall `POST /v1/recall` takes a query and an end-user id, returns ranked memories. Drop them at the top of your prompt — that's it.
3.	Multi-tenant out of the box — built for B2B from day one workspaces, API keys, per-tenant rate limits, audit log, end-user scoping. Your customers' users are isolated by default.
4.	Vector search, no Pinecone — powered by Cloudflare Vectorize 768-dim embeddings via Workers AI, namespaced per workspace. Sub-100ms recall at the edge. No infra to provision.
5.	Consolidation handles the messy parts — conflicts, duplicates, decay when memories contradict ("lives in NYC" vs "lives in SF"), we resolve and version. Old low-confidence memories decay nightly.
6.	PII detection, GDPR endpoints — for the compliance review email / phone / card detection on every write. One-call data export and delete per end-user.

Pricing

Hacker	free	10k memories, 1 workspace
Startup	$29/mo	500k memories, 5 workspaces
Scale	$199/mo	unlimited, audit export
Enterprise	contact	SOC2, BYOK, SSO

all tiers include unlimited end-users.

Quick start

curl -X POST https://api.northcontext.dev/v1/memories \
  -H "Authorization: Bearer $NC_KEY" \
  -d '{
    "end_user_id": "user_42",
    "text": "I work at Acme as a senior PM, prefer slack over email."
  }'

Who's using it

(showcase placeholder — add logos here once you have design partners.)

Acme Support · ChatStack · SidekickAI · PromptCraft · Loop · Verdant