The product
LocalBrain: one governed path between AI and your data.
LocalBrain is a local-first control layer. Whether the request comes from the built-in workspace, an external AI client over MCP, or a custom app over HTTP, it goes through the same governance: identity, policy, approval, and evidence — before any protected action can happen.
Capabilities
What LocalBrain does
Sensitive-data detection
Classifies data before processing — PII, PHI, secrets and organization-specific terms via installable privacy packs.
Local vs. cloud routing
Policy decides where each request runs. Sensitive content stays local or is masked before any cloud call.
Least-privilege agents
Read, write, modify, execute and network are each granted explicitly. Governed actions are default-deny.
Human approval
Plan → approve → act. Risky steps pause for a person; nothing executes before approval.
Tamper-evident evidence
A hash-chained record of who did what, under which policy, with what result — verifiable and exportable.
Any model, one gateway
Local, private or cloud models — including cheaper open-weight models — all run under the same controls.
External-agent gateway
External AI clients connect over MCP or an authenticated HTTP API and see only governed tools — never your raw credentials.
Credential protection
Agents get a reference, not the secret. Downstream credentials are resolved only at approved execution time.
Runs offline
Deploy on a workstation or an organization-controlled server. Nothing leaves your environment unless policy allows it.
The governed flow
Every request takes the same path
The request and its documents are scanned. Sensitive information is detected and labelled before anything runs.
Policy chooses local or cloud for this request. Sensitive content is kept on-device or masked first.
The action is checked against least-privilege permissions and organization policy. Unmatched actions are denied.
If the action is consequential, it pauses for a human. The exact intent is what gets approved — no substitution.
Only an approved action runs, through a trusted executor that never hands the agent your raw credentials.
The classification, route, decision, approver and result are written to a tamper-evident record you can verify.
Editions
One product, three tiers
Start local, grow into the organization. The same governed core runs across all three.
| Edition | Who it's for | What's included |
|---|---|---|
| Basic | Individuals & small teams | Local-first document & chat workspace, local models, PII detection & local-only rules, offline operation. |
| Pro | Consultants & regulated-industry pilots | Everything in Basic, plus the pilot workspace, governed document-agent flow, permissions, human approval, DPG controls, evidence & audit, pilot report export. |
| Enterprise | Healthcare / pharma organizations | Everything in Pro, plus organization & role policy hierarchy, identity/SSO-ready, credential broker, SIEM/webhook export, admin console, secure organization-hosted access. |
Availability and packaging are confirmed per engagement. Ask us what fits your environment.
Put it against a real workflow
We'll set up a pilot on non-production data and show you the governance end to end.