Documentation

Everything, organized.

Most docs hubs make you guess where the answer is. This one names the destination on every tile.

Start here

30 min
30-minute onboarding runbook →
Minute-by-minute: sign up, install SDK, insert bitemporal facts, backdate a change, record agent memory, connect Claude Desktop via MCP. Every command tested against tonight's prod binary; failure output printed next to the fix.
5 min
Quickstart →
Three paths: AI agent via MCP, Python/Node SDK, or curl against the HTTP API. Copy-paste configs, working snippets, common errors.
Reference
Features →
Complete catalog of every shipped primitive: storage, Cypher, search, graph algos, agent memory, compliance, server, auth, SDK, MCP, replication, ops.
Guide
AI memory →
Episodic log, belief revision, salience decay, working-memory consolidation, conversation sub-tenancy, vector+graph fusion. End-to-end Python pattern.
Verticals
Regulated industries →
Healthcare, finance, insurance, public sector. HIPAA — GDPR — SEC 17a-4 — NAIC — FedRAMP. Each compliance question mapped to a primitive.

Reference docs

14 tools
MCP tools →
Full catalog of the Model Context Protocol tool surface. cypher_query, search, agent-memory primitives, compliance procedures. Claude Desktop / Cursor / VS Code config.
Python + Node
SDK reference →
Idiomatic clients for Python and Node. Cypher, search, indexes, cursor streaming, capability + Cognito tokens.
Language
Cypher reference →
openCypher core + bitemporal extensions (AT VALID / AT RECORDED / OVER VALID BETWEEN). Procedures, functions, gotchas.
OpenAPI
HTTP API spec →
Interactive Swagger UI for the full HTTP surface. Every endpoint typed; every response schema documented; try-it-now panel.
Operator
Compliance dashboard →
Runtime compliance dashboard: audit chain status, ACL events log, crypto-shred bundle export, WORM mode toggle, DP budget view.
Design
Design principles →
The capability-level guarantees InvariantDB commits to. Each principle describes what the engine promises callers, not how it's built inside.

By role

If you're building an agent →
Connect your AI via MCP (30 seconds), then read the memory guide. Replay any past decision; isolate per-conversation.
If you're in procurement →
Start with regulated industries, then compliance. Every claim reduces to a design principle.
If you're a developer →
Quickstart → SDK reference → Cypher reference → HTTP API spec. Build something.
If you're evaluating →
Full feature catalog with shipped-vs-deferred honesty section. Or watch six demos on the homepage.

Operate it

Runbook
Operations →
Liveness, deploys, rollback, backup, high availability, incident response — the customer-facing shape of operational guarantees.
DR
Backup · restore · DR →
Portable, cryptographically verifiable backups. Online capture, offline restore, integrity refuse on drift.
Cluster
High availability →
Strongly-consistent multi-node replication with automatic failover; reader scale-out; membership changes.

Help, support, talk to us

The fastest way to reach us is the signup form — it captures who you are and what you're trying to build, and we'll reply with a personal note from the founder. NDA-protected conversations are fine.

For deeper questions: every page on this site links to the underlying capability. If the docs don't answer it, we want to know — that's the gap we'll fill next.

Looking for the legacy v1 docs? They live at docs-v1.html. The four pillars above (Quickstart, Features, AI memory, Regulated industries) cover everything in v1 plus the 2026 ship.