Govern AI agents on MCP Bridge
Add governance to MCP tool calls via a bridge that sits between MCP clients and servers. The MCP Bridge pack evaluates every tool invocation against policy before forwarding to the target server.
What this pack does
- Pre-dispatch policy evaluation on MCP tool calls
- Gateway-level approval gates
- Output safety checks on tool responses
- Protocol-native audit trail
Use cases
Govern all MCP tool calls from Claude, Cursor, or custom clients
Require approval for file system and database MCP tools
Audit every MCP interaction across your organization
Quick setup
- 1Install the MCP Bridge pack: cordumctl pack install mcp-bridge
- 2Configure upstream MCP server endpoints
- 3Define tool-level governance policies
- 4Point your MCP clients to the Cordum bridge endpoint
Frequently asked questions
Where does the MCP Bridge sit in the request path?
The bridge sits between your MCP clients and upstream MCP servers. It submits each MCP tool call as a Cordum job so the Safety Kernel evaluates it before the call is forwarded, making every tool invocation policy-gated at the gateway.
Which MCP clients can route through the bridge?
Any MCP client works, including Claude, Cursor, and custom clients. You point them at the Cordum bridge endpoint instead of the upstream server directly, and governance applies uniformly across every connected client.
Can I require approval for specific MCP tools?
Yes. Define tool-level policy so high-risk tools, such as file system or database tools, return a Require Approval decision before the bridge forwards the call. Write-style actions are gated while read-style tools can be allowed to pass through.
What does the audit trail capture for MCP traffic?
Every MCP interaction is recorded with a protocol-native audit trail that captures the tool call, the governance decision, and any approval. That gives you organization-wide visibility into MCP tool use across all of your clients and servers.
Ready to govern MCP Bridge?
Other integrations
Approval notifications and agent alerts in Slack channels.
Govern AI agent actions on GitHub repositories.
Govern AI agent read and write actions against the AWS API.
Governance for AI agents managing Jira workflows.
Govern AI agents responding to Kubernetes incidents.
Feed Datadog alerts into governed agent workflows.