Read the AccessGraph
Connect agent actions to business-critical asset classes, required controls, and the local Guardprint drift signal.
Docs
How CapitalGuard works.
Practical documentation for customers, operators, affiliates, and developers implementing the live service.
Use prevention playbooks
Turn scanner and report findings into package-specific action packets, evidence lists, and controls to install.
Authorize a scan scope
Name the repository, confirm ownership or written authorization, and define the exact defensive scope before review.
Run the local scanner
Inspect common AI-agent exposure paths inside the repository without uploading source code or printing secret values.
Read a report
Every finding explains severity, affected path, agent exposure, business impact, preventive control, and confidence.
Generate policy
Create guardrails for Codex, Claude Code, Cursor, Copilot, GitHub automation, or internal LLM agents.
Activate licenses
Use the private key with the matching purchase email to activate the delivery vault and purchased scan allowance.
Retrieve private files
Download the installer, policy starter, safe-use guide, certificate, checklist, and delivery terms from the protected vault.
Export evidence
Use executive PDF reports for investors, customers, procurement, and internal security review.
Submit scan intake
Spend one purchased credit only after confirming the exact repository scope and your authority to request the review.
Affiliate engine
Approved creators receive links, campaign codes, assets, commission visibility, and payout status.
Developer integration
GitHub Actions AI-agent security gate
Install the cited eight-control profile and immutable starter workflow.
Reproducible evidence
GitHub Actions AI-agent gate benchmark
Inspect ten inert workflow fixtures, eight control predicates, exact results, and release digests.
Standards evidence
OWASP Agentic Top 10 evidence map
Trace all ten risk areas to versioned control releases, runtime evidence, methodology, and SHA-256 source bindings.
Signed license evidence
AI Agent Security License verification benchmark
Reproduce one accepted signed chain and eight authority, expiry, revocation, binding, and schema rejections.
Customer workflow
Choose one of three one-time licenses
Complete payment in hosted Stripe checkout
Receive the private license and customer kit
Activate with the matching purchase email
Confirm ownership or written authorization
Run the local exposure scanner and guardrail installer
Submit the purchased scan scope
Receive the report, policy artifacts, and approved proof path