The federated enterprise knowledge platform you can actually host yourself.
Moʻolelo is a self-hosted knowledge substrate for organizations that can't put their data in someone else's cloud — unified retrieval, knowledge graph, and supersession over your wikis, repos, and document libraries, built around subject-level data residency, source-system ACL fidelity, and first-class air-gap deployment.
moʻolelo — the Hawaiian word for narrative, history, accumulated stories — including the things that are not written down.
Architectural commitments
Three things we won't compromise.
These aren't features we'll add later — they're the reason the platform exists. Whether you're running it on a single node for a small team or across regions for a multinational with hundreds of thousands of employees and a petabyte-class corpus, you get the same architectural commitments.
Subject-level data residency
Personal data of any individual physically resides only in the regional vault for that individual's residency. The global graph carries opaque tokens, never PII. Queries federate across regions; PII never does.
Source-system authority is canonical
Moʻolelo never grants access the source system wouldn't grant. It can be more restrictive, never less. Every authorization decision is logged with the source authority chain.
Air-gap is first-class
No SaaS in the data plane. No internet-connectivity requirements at runtime. Every container image, Helm chart, ML model, and dependency mirrors to your registry.
Knowledge intelligence
Beyond search: the corpus that learns about itself.
A search index tells you what's in the corpus. A knowledge platform tells you what's missing from it — and where the institutional knowledge that should be written down is hiding. These are the surfaces the substrate is being built to support. None ship today; all are in active design.
Conversation ingest
Slack threads, Teams channels, meeting recordings, and LLM-agent traces (Langfuse, OpenLLMetry) flow into the same residency-clean substrate as your wikis and code — so institutional memory captures the things that are not written down, not just the things that are.
Knowledge-gap intelligence
The unified retrieval + graph + query-log signal surfaces where the corpus is thin: topics queried frequently with no good answer, entities heavily referenced but never written about, areas where a key contributor's expertise lives only in conversations. The output is "write more about X here" recommendations directed at the right author — not synthetic content.
Learning loop
Every served result, every graph expansion, every supersession decision is already audit-stamped today. The next step is closing the loop: feeding which results actually answered questions back into ranking, supersession authority, and reference confidence — without sending any of that signal outside your residency boundary.
We don't pre-sell features. These are listed so you can see where the substrate is heading and decide whether the direction matches your problem — not so you can buy them today. Each ships behind the same architectural commitments: subject-level residency, source-system authority, air-gap deployment.
Honest fit
When Moʻolelo is the right answer.
Size doesn't decide it — what you're trying to do does. The same platform serves a five-person team running it on a single node and a hundreds-of-thousands-employee multinational running it across half a dozen regions over a petabyte-class corpus. The question is whether your problem shape lines up with what Moʻolelo actually solves.
A good fit when
- You're running on your own infrastructure and don't want a SaaS-only enterprise knowledge platform
- You need search, graph, and supersession results that exactly honor source-system permissions — no over-grants, no under-grants
- You care where personal data physically resides and won't pool it across regions
- You're building agentic workflows or LLM agents that need grounded organizational context — both the documents AND the relationships between them
- You want a platform you can read, audit, and modify — not a black box
Probably not the right fit when
- You want a managed multi-tenant SaaS — Moʻolelo isn't one and won't be
- You want a chat surface as the product — Moʻolelo is the knowledge substrate, build the chat on top
- You don't have any source-system identity authority (HRIS, IdP, or equivalent) — we can't do residency or ACLs without it
Design partner program
Get early access.
We're onboarding a small number of design partners. Tell us about your environment and we'll get back to you within two business days.