Domain objects
The customers, accounts, cases, claims, deliverables, decisions, or workflows your product is actually about.
Build on Penumbra
Build agents on a real domain model.
Describe the objects, rules, workflows, and standards your product needs. Penumbra turns that domain model into tools, APIs, extraction, memory, guardrails, and provenance your agents can use.
What the domain model gives you
The domain model becomes the working parts of the AI system.
Agent tools
Tools generated around the domain model, so agents can act on real business objects instead of loose text.
APIs
Typed interfaces over the domain model, so your app and agents work from the same definition of the domain.
Extraction
Documents, calls, tickets, and systems hydrate the domain model directly instead of producing another blob of text.
Memory
Work leaves behind structured state agents can reuse, not just transcripts someone has to search later.
Guardrails
Rules, review standards, and permissions travel with the domain model instead of living in scattered prompts.
Provenance
Every output can show what source, rule, domain object, or prior decision shaped it.
Fast context
Agents get the relevant domain context and state at runtime without carrying the whole company in the prompt.
The problem
Every AI tool is making up its own version of the business.
The vector store knows documents. The workflow engine knows steps. The database knows tables. The agent knows prompts. None of them knows the domain the way your product needs it.
Penumbra gives the stack one domain model. Agents can use it to find context, call tools, follow rules, remember work, and explain where outputs came from.
How to adopt it
Start with one useful slice.
Start with the thing you are building
Map one workflow, agent, product surface, or corpus. You do not need an enterprise-wide modeling program to start.
Keep meaning in one place
Without a shared domain model, every tool in the stack invents its own version of the domain. Penumbra gives them one source.
Let agents act, not just answer
Agents need tools, memory, rules, and provenance. Penumbra turns the domain model into the pieces they need to do work.
Grow the domain model as the product grows
Each workflow adds more structure. The company brain emerges from connected useful pieces, not one giant upfront schema.
The simple version
Give AI the domain before you ask it to do the work.
A prompt can explain a task. A domain model lets the system keep working after the first answer: with tools, memory, rules, and provenance.
Bring us a workflow →