MCP Server Mesh
View and manage the registered MCP server estate.
MCP Server Mesh is the inventory surface for every MCP server a tenant has registered — as opposed to MCP Server Connection, which is the form that adds or edits one. The root page (/) lists every server as a row: a type badge, name, endpoint, a live tools/resources count pulled by grouping the catalog's tool and resource rows by mcp_server_id, a policy-mode pill (observe, simulate, or enforce), a health dot, and the last discovery sync time. Search and status/type filters narrow the table client-side; there is no server-side pagination today.
Each server's row links to /servers/:id, a dedicated detail page with a graph tab (the server's tools and resources rendered as a filtered slice of the AI Graph), a catalog tab (its tools, resources, tables, and prompts), a policies tab (every published policy attached to the server itself or to one of its tools/resources, bucketed by target), an actors tab (which AI Actors have called through this server), and a decisions tab (the same enforcement-log table used by the standalone Decisions Log, pre-filtered to this server). A "Discover Now" button on the detail page triggers an out-of-cycle discovery pass rather than waiting for the scheduler's next tick.
"Governance state" is read off the server, not computed by this capability: policy mode (observe / simulate / enforce) and health status are both fields on the underlying gcdm_mcp_servers row, written by the policy and health-probe subsystems respectively — the Mesh page just displays them alongside the count of currently-enforcing servers in the page header.
Reference
Features
- description
- The full list of registered MCP servers, with type badge, endpoint, live tools/resources counts, policy mode, health, and last sync time. Searchable by name, endpoint, or type; filterable by status and type.
- description
- The stored attributes of a server row: vendor, environment, endpoint URI, transport type, auth type, protocol version, and its declared MCP capabilities.
- description
- The health dot shown per server (healthy, degraded, unreachable, or unknown), sourced from the periodic health probe rather than computed on this page.
- description
- The per-server page at /servers/:id: summary cards (tool/resource counts, read vs. write split, native-view availability, last sync), then tabs for graph, catalog, policies, actors, and decisions.
- description
- The Catalog tab on a server's detail page — its discovered tools, resources, tables, and prompt templates, scoped to that one server.
- description
- The Actors tab on a server's detail page, listing which AI Actors have invoked tools or resources on this server.
- description
- A server's policy mode (observe, simulate, or enforce) and health status, both stored on the server row and surfaced as pills/dots in the inventory table and detail header.

Docs