OWNSTATE
For institutions ↗
Move without rebuilding the brief.
A founder, researcher, engineer, or operator carries project decisions, artifacts, and useful context between the models and workspaces they choose.
What your institution knows should outlive every tool.
Ownstate is the independent state layer beneath your AI stack. It preserves evidence, builds versioned institutional knowledge, and supplies authorized context to the people, models, agents, and applications that need it.
Institutional knowledge that outlives models
AI work creates valuable understanding. Too often, it is left behind.
Decisions, research, failed approaches, and commitments become trapped in old chats, agent sessions, and vendor memory. Teams re-brief. Agents repeat work. The reasoning behind important choices disappears. Ownstate separates state from intelligence: models think; your durable record stays yours.
Sovereign state infrastructure for AI work
One memory. Every intelligence.
Ownstate is the independent state layer beneath the systems you choose. Swap a model, invite a specialist agent, or coordinate a fleet without abandoning the knowledge that makes their work matter.
Preferred modelSpecialist agentsAgent fleet
OWNSTATE MCPGoverned context.
Enduring knowledge.
Enduring knowledge.
Changing a model changes the intelligence at the edge. It does not change the institutional knowledge beneath it.
The context contract
The model receives a bounded context. You retain the state.
Each request is assembled for the work at hand, not copied from an opaque chat history. Scope, authorization, evidence, and provenance travel together.
- 01Relevant by design
Only task-relevant records and project context are compiled into the packet.
- 02Governed before delivery
Authorization and classification policies are applied before context leaves your boundary.
- 03Inspectably grounded
Important knowledge remains connected to its evidence and recorded provenance.
Evidence, canonical state, query, and context
MCP is one interface into Ownstate—not the product itself.
- 01 / CONNECT
Add Ownstate MCP
Connect the canonical MCP server to the model or agent environment you choose.
- 02 / BOOTSTRAP
Start with context
At session start, the agent receives a scoped project brief and the knowledge relevant to its task.
- 03 / WORK
Retrieve with controls
Search and context compilation apply authorization, classification, and provenance before delivery.
- 04 / PRESERVE
Keep what matters
New work is recorded as evidence; candidate knowledge is validated before it becomes canonical.
Governed context for people, models, and agents
The next model should know what the last one learned.
Turn past work into shared context.
Give people a grounded view of requirements, decisions, risks, and rationale without relying on one person’s memory or one model’s chat history.
Give agents a common ground.
Research, finance, engineering, and legal agents can contribute from the same institutional state, each scoped to only the context they are allowed to use.
From the work already done
The next model should not have to begin at zero.
Ownstate makes prior work usable rather than buried. Ask the questions that normally trigger another search through old chats, documents, and people.
- 01What did we decide, and why?
- 02What work has already been done on this?
- 03What changed in the last six months?
- 04Which evidence supports this answer?
Sovereignty is a deployment choice
Run where your institution needs its knowledge to live.
Control is the feature
Institutional memory should not be a black box.
Provenance, not guesswork
Every important record can retain the evidence that supports it, so answers are inspectable, not merely plausible.
History, never overwrite
Canonical knowledge is versioned. What changed, when it changed, and why it changed remains visible.
Authorization before intelligence
Policy and classification shape retrieval and model context before information ever leaves your boundary.
Independence by architecture
Models, agents, vendors, and interfaces remain replaceable. Your organization’s accumulated knowledge does not.
For institutions that need continuity
Keep the state.
Change the intelligence.
Every decision, artifact, and agent session should make your institution more capable. It should not vanish when the tool changes.
- •Carry context across models, teams, and agent environments
- •Keep evidence beneath the knowledge that guides work
- •Give every agent enough context, not everything
Explore the project on GitHub.