company architecture · 2026-05

One substrate.
Three products.

InvariantDB is the temporal-truth runtime. InvariantDB is the AI memory platform built on it. StaticOwl is the deterministic publishing platform built on it. Three distinct surfaces, three buyer profiles — one engineering team, one engine.

01The shape

substrate · the engine
InvariantDB
Bitemporal graph runtime. Tamper-evident audit chain. Provable invariants. The database for systems that must explain themselves.
flagship product
InvariantDB
Sold directly as the database for regulated AI workloads. Receipts for every action your agents take, evaluable against the policy in force at the time.
platform
InvariantDB
AI memory + knowledge platform. Long-term memory for LLM agents, temporal context, MCP integrations, semantic + graph retrieval.
platform
StaticOwl
Graph-authored publishing + CMS. Deterministic builds with full dependency lineage, time-travel previews, explainable rebuilds.

The substrate brand and the flagship product brand are the same — InvariantDB — because compliance buyers are looking for "the database that proves things," and that's the substrate sold as itself. InvariantDB and StaticOwl are opinionated platforms that build on the substrate for different audiences.

02Why this hierarchy

The three pieces have radically different buyers, sales motions, and price points. Forcing them under one brand creates marketing mush. Splitting them under one substrate creates a coherent ecosystem.

InvariantDB

the database · invariantdb.com

A bitemporal graph runtime with a tamper-evident audit chain. Every fact has both valid time and transaction time; every mutation is signed into a SHA-256 chain; every primitive (sort, merge, audit-link, span correctness) has a formal proof in the invariant catalog. Cypher-queryable, Raft-replicated, formally verified where it matters.

Tagline "Temporal truth infrastructure for systems that must explain themselves."
Buyer CISO, Chief Compliance Officer, regulated-industry infra lead
Sales motion Direct, top-down, multi-month cycle
Pricing Enterprise contracts ($100K+ ACV) + free OSS engine tier
Use case anchor AI agent receipts in regulated industries (insurance, healthcare, finance)
Differentiator Policy evaluation against historical state, not just current state

InvariantDB

memory + knowledge platform · graphiquity.com

Long-term memory for LLM agents. Bitemporal, graph-shaped, MCP-native. Devs sign up, install the SDK, and start writing/reading agent memories in minutes. Beneath the memory.remember() / memory.recall() API is a multi-tenant InvariantDB cluster — but the developer experience hides the runtime.

Tagline "Memory for agents that remembers when, not just what."
Buyer Developer building an AI agent or assistant
Sales motion Self-serve, bottoms-up, sign-up to first call in minutes
Pricing Free tier + $20–200/mo team + custom enterprise
Use case anchor Cross-session agent memory; "as we knew it" recall
Differentiator vs Mem0 / Letta / Zep Bitemporal recall, tamper-evident chain, reverse queries (who knows about X?)

StaticOwl

graph-authored publishing

A static-site / content engine where the source is a graph, not a folder of Markdown. Deterministic compiles from the graph state, lineage-tracked rebuilds, time-travel previews, explainable dependency invalidation. Content teams author in StaticOwl; the InvariantDB substrate gives every publish a signed receipt.

Tagline "Publishing with provenance."
Buyer Content / marketing teams, developer-relations editors
Sales motion Self-serve to team-tier upgrades
Pricing Free tier + team plans
Use case anchor Graph-modeled docs, knowledge bases, marketing sites with explainable rebuilds
Differentiator Time-travel preview ("show me the site as of last Tuesday"); dependency lineage in the graph itself

03How they reinforce each other

The three brands aren't fenced off from each other — they share an engine, and that creates real adoption flywheels.

graphiquity adoption ──→ engineer at regulated company uses InvariantDB Memory │ └→ their CIO/CISO later asks: "what AI is in our stack?" │ └→ engineering team answers: "we use InvariantDB, powered by InvariantDB" │ └→ compliance buys InvariantDB receipts on top of the same engine │ └→ procurement is short-circuited (engine already in the stack) │ [bottoms-up + top-down · landing-and-expand] staticowl adoption ─→ marketing team uses StaticOwl for docs │ └→ adjacent engineering team sees the graph-substrate pattern │ └→ adopts InvariantDB for their own provenance needs │ [horizontal expansion within the customer]

The pattern is what worked for HashiCorp (one runtime, many opinionated products), MongoDB (engine + Atlas), Elastic (engine + Kibana). One foundation, many surfaces, multiple buyers.

04Brand voice — who talks to whom

The three surfaces speak in different registers. Same engineering integrity, different posture.

InvariantDB

Substrate gravitas. Compliance language. Whitepapers, SOC 2 attestations, analyst briefings. Voice: "the database that proves things."

InvariantDB

Developer energy. Technical-but-accessible. Docs-first, conference talks, GitHub. Voice: "memory that remembers when, not just what."

StaticOwl

Tooling clarity. Editor-friendly. Visual diffs, build-output streams. Voice: "publishing you can audit."

None of the three are "the playful one." All three carry the substrate's integrity. They differ in register, not seriousness.

05What this commits us to