GeoClear Documentation.
Build with operational evidence for AI actions, tool calls, workflow handoffs, and customer-held verification.
GeoClear helps receiving systems verify whether an action followed the approved evidence path before accepting it. These docs introduce the verification layer, operational evidence records, customer-held evidence bundles, local verification, and deployment patterns.
Start here.
Six core concepts behind the operational evidence layer.
What is operational evidence?
An operational evidence record that travels with an action, carrying actor / action / policy reference / commitments / verification result. The customer keeps it.
How the verification layer works
The interlock sits in front of tool calls. Valid evidence proceeds. Missing evidence holds. Policy violations block. Tampered evidence rejects.
What gets sent
An operational evidence record. Commitments and verification material travel with the action; raw source data does not.
What stays with the customer
Raw mission data, PII, raw imagery, invoices, internal logs, and full policy systems stay inside the customer boundary by default.
Where policy runs
GeoClear supports multiple deployment postures for where the policy result is produced. Specific deployment-mode details are shared with qualified evaluators under evaluation access.
How local verification works
The receiving system uses customer-held verification material to check the operational evidence record without a live GeoClear application call.
Build paths.
Five integration shapes for adopting operational evidence in production.
Agent tool calls / MCP
MCP-compatible middleware over standard transports. Sits in front of tool calls or server-side handlers. Structured guidance returned on hold / reject. See Verification for MCP.
API gateway / workflow interlock
Same evidence packet shape, gateway integration pattern. Sits as middleware in front of workflow handoffs and API surfaces. Architecture brief on request.
Customer-held evidence bundle
Portable evidence bundle, such as a ZIP or equivalent customer-held package, carrying the operational evidence record, payload summary, verification material, and an offline verifier. Customer retains the artifact for later review.
Local / offline verification
The receiving system verifies locally using customer-held verification material. No live GeoClear application call required at verify time. See the offline verifier.
Customer-controlled deployment
GeoClear software runs inside a customer-controlled or authorized integrator-managed environment for sovereign, federal, DDIL, and high-assurance workloads.
Policy Foundry.
Controlled-preview policy intake for supported policy families.
GeoClear supports controlled-preview policy intake for supported policy families. A customer can express policy in plain English; GeoClear structures it into an evidence contract; ambiguous cases route to review; runtime systems use the contract to accept, hold, block, reject, or escalate actions.
Agent Harness.
Plan-Execute-Verify discipline for AI agent behavior at deploy time.
GeoClear applies Plan-Execute-Verify discipline to AI agent behavior at deploy time. Behavioral checks gate the CI/CD release. The customer-held operational evidence record verifies what was tested. See Agent Harness for CI/CD.
AUTH.md customer-repo policy file.
A repository-level policy input treated as policy context, not as model authority.
AUTH.md is a repository-level policy input. It can describe allowed actions, prohibited actions, approval requirements, model and service-card requirements, evidence requirements, and escalation rules. GeoClear converts supported policy families into structured evidence contracts and routes ambiguous policy to review rather than fabricating precision. GeoClear treats AUTH.md as policy context, not as model authority.
What gets sent. What stays with the customer.
The trust-boundary contract in two lists.
Sent with the action
- Actor
- Action
- Policy reference
- Evidence commitments
- Approval state
- Freshness
- Signature / issuer reference
- Verification result
- Customer evidence record
Stays with the customer
- Raw imagery
- Raw logs
- PII
- Invoices
- Mission data
- Full policy systems
- Source systems
Where policy runs.
Multiple deployment postures. One verification path.
GeoClear supports multiple deployment postures for where the policy result is produced, matched to commercial, sovereign, DDIL, and federal requirements. Specific deployment-mode details, selection guidance, and integration specifics are shared with qualified evaluators under NDA or through approved portal access.
Private evaluator materials.
Detailed schema, verifier internals, key-management posture, package instructions, and deployment-specific integration guides for qualified evaluators.
Need the exact integration guide?
Exact schema, verifier internals, key-management guidance, package instructions, and deployment-specific details are available to qualified evaluators under NDA or through approved portal access.
The neutral operational evidence layer for AI actions.
GeoClear is not a location API with operational evidence records. GeoClear is operational evidence infrastructure for AI actions, tool calls, and workflow decisions.
GeoClear issues operational evidence records for AI actions, tool calls, and workflow decisions. The evidence bundle is tamper-evident, and verifiers detect modification. GeoClear does not provide legal certification and does not verify the real-world truth of upstream data sources.