Documented Product Controls
The following descriptions are limited to capabilities documented on AgentGuard’s public site. They do not establish certification, compliance, or complete coverage.
Security
AgentGuard Security and Data Boundaries
Review the documented controls AgentGuard applies to agent actions and components, what public sources say about local and connected data handling, and where current integration limits remain.
CTRL-01Documented Product Controlsdocumented
CTRL-02The following descriptions are limited to capabilities documented on AgentGuard’s public site. They do not establish certification, compliance, or complete coverage.documented
CTRL-03Runtime Guarddocumented
CTRL-04Runtime Guard evaluates security-relevant actions before execution, including shell commands, file access, tool actions, network requests, secret access, sensitive writes, and potential webhook exfiltration.documented
SECURITY / 01
Documented Product Controls
Runtime Guard
Runtime Guard evaluates security-relevant actions before execution, including shell commands, file access, tool actions, network requests, secret access, sensitive writes, and potential webhook exfiltration.
Explore Runtime Guard
Deep Scan
Deep Scan reviews skills, plugins, MCP servers, agents, and related components for documented risks such as prompt injection, malicious tools, credential leaks, and backdoors.
Explore Deep Scan
SECURITY / 02
Local and Cloud-Connected Data Handling
Local and Cloud-Connected Data Handling
Local mode
AgentGuard’s public FAQ states that local mode does not upload full code, prompts, secrets, or file contents.
Cloud-connected use
Public sources state that connected use may send sanitized action previews, risk metadata, decisions, policy versions, redacted metadata, and audit events when needed.
Exact processing depends on configuration and use. Review the approved Privacy Policy and any applicable agreement before relying on these statements for a legal or procurement decision.
Read the Privacy Policy
SECURITY / 03
Integration Depth and MCP Limits
AgentGuard documents different integration modes across supported agent and IDE environments. Some use pre- and post-tool hooks, some use plugin hooks, and others use skill- or command-based paths. Coverage should not be assumed to be identical.
AgentGuard’s public FAQ states that it cannot currently fully monitor or block every third-party MCP server runtime call. MCP server scans, reputation, trust-registry signals, and available hook layers can still support risk decisions within their documented scope.
Review MCP Security
SECURITY / 04
Security Evidence and Assurance Materials
Not yet published
Security Evidence and Assurance Materials
Use this section only for current, owner-approved evidence. Product documentation is available today; the remaining materials must not appear as badges or claims until verified.
Product security documentation
Available through AgentGuard Docs
Third-party security assessments
Not yet published
Security certifications
Not yet published
Encryption and key-management details
Not yet published
Data retention and residency information
Not yet published
Subprocessor information
Not yet published
Vulnerability disclosure channel
Not yet published
Request Available Security Information
SECURITY / 05
Policies and Security Contact
Policies and Security Contact
Privacy Policy
Review the Legal-approved policy that governs AgentGuard privacy disclosures.
Read Privacy Policy
Terms of Service
Review the Legal-approved terms that govern use of AgentGuard services.
Read Terms
Security questions
Send your environment, control, or procurement questions without including credentials, source code, API keys, or production data.
Contact Security
SECURITY / 06
Security Questions
Security Questions
Review the documented evidence and current product boundary for this requirement.
Does local mode upload full code, prompts, secrets, or file contents?
AgentGuard’s public FAQ states that local mode does not upload those full contents. Configuration and connected features still require review against the approved Privacy Policy.
What may connected use send?
Public sources describe sanitized action previews, risk metadata, decisions, policy versions, redacted metadata, and audit events as data that may be sent when needed.
Can AgentGuard block every third-party MCP runtime call?
No. The public FAQ states that complete monitoring or blocking of every third-party MCP server runtime call is not currently supported.
Which certifications or audit reports are available?
Only materials listed in the owner-approved evidence directory should be treated as available. No certification or audit status is asserted by this page structure.
SECURITY / 07
Review AgentGuard for Your Environment
Review AgentGuard for Your Environment
Map your agent hosts, tool actions, components, data boundaries, and residual MCP risks before deciding which controls fit your workflow.
Contact Security
Open AgentGuard Docs
official_homepage
official_docs_quickstart
context/product-fact-base.md
Review the Security Boundary
Map documented controls, data handling, integration depth, and residual risk to your environment.