01
Native tool access
Make AgentGuard available through the tool model already supported by an MCP-capable host.
Use AgentGuard In MCP Workflows
Expose AgentGuard as an MCP server so a compatible host can request security checks through its native tool interface.
The MCP Server method makes AgentGuard callable. It does not establish complete monitoring or blocking of every third-party MCP runtime call.
MCPHost-invoked checks
MCP ServerIntegration method
AgentGuardSecurity service
ConnectReviewEvidence
ConnectionMCP Server
Service behaviorHost-invoked checks
Delivery modelNative MCP tool access
Current statusClaude example available
01 / CONNECTION
How AgentGuard Connects
See which supported AgentGuard forms apply to this target and what connection work still remains.
02
Server configuration
Keep one named AgentGuard server entry and review the tools and permissions exposed to the host.
03
Host verification
Confirm discovery, approval behavior, invocation, and returned evidence for each supported MCP host.
04
Runtime boundary
Record host actions and third-party MCP calls that do not invoke AgentGuard.
02 / CONNECT
Connect, Configure, Verify
Commands match the current public setup paths. Production documentation should add approved versions, prerequisites, rollback steps, and source verification.
MCP · current verified runtime commands
# Add AgentGuard as an MCP server in Claude Code
claude mcp add agentguard -- npx -y @anthropic-ai/mcp-remote@latest https://github.com/GoPlusSecurity/agentguard
# For other MCP hosts
# Use the approved host-specific configuration and verify tool discoveryAdd the MCP server
Use the current public Claude Code example or the equivalent approved command for another tested host.
Inspect tools
Confirm AgentGuard appears in the host and review the security tools made available.
Test one request
Invoke a controlled check and document the result and calls outside the AgentGuard path.
03 / SURFACE
Integration Surface
These labels separate installation, service access, workflow connection, and runtime behavior.
Integration formMCP Server
Public host exampleClaude Code
Other MCP hostsVerify separately
Agent accessCallable tools
Automatic interceptionNot established
Third-party callsCoverage limited
04 / EVIDENCE
Verify The Connection And Result
Check the connection, runtime, returned result, and coverage boundary before treating setup as verified.
The MCP host lists AgentGuard as an enabled server.
The expected AgentGuard tools and descriptions are visible.
A controlled security request returns reviewable output through the MCP host.
MCP calls that never invoke AgentGuard remain outside this service path.
05 / LIMITS
Known Limitations Stay Visible
These statements prevent a documented integration from being read as universal protection.
MCP setup, permissions, tool discovery, and approvals differ by host.
An available MCP server does not prove the host will call it for every relevant action.
Do not claim complete monitoring or blocking of other MCP server calls.
Publish tested hosts and versions before expanding beyond the current Claude Code example.
06 / FAQ
Integration Questions
Keep answers aligned with current product behavior and the last verified integration version.
Is this page about scanning MCP servers?+
This integration page focuses on using AgentGuard itself as an MCP server. MCP risk and server scanning belong on the MCP Security solution page.
Which host has a public command?+
The current homepage provides a Claude Code MCP command.
Does the MCP method cover every tool call?+
Coverage follows the AgentGuard tools the host and agent actually invoke.
MCP integration
Verify The Integration Before Production
Run the documented setup, confirm AgentGuard is reachable from MCP, and test the behavior expected from this connection.