MCP page shows "Not available" — workspace calls /api/mcp instead of /api/mcp/servers

[sami-deepguards]

Environment:

• Hermes Agent v0.19.0 (2026.7.20), docker install
• hermes-workspace via Docker Compose (ghcr.io/outsourc-e/hermes-workspace:latest)
• Both services on network_mode: host, dashboard bound to 127.0.0.1:9119 (loopback, no auth provider configured)
• workspace-overrides.json: claudeApiUrl: http://127.0.0.1:8642, claudeDashboardUrl: http://127.0.0.1:9119

Bug:
The MCP page in the workspace UI shows "Not available on this backend. Connect to a Hermes Agent gateway to unlock MCP Servers," even though MCP is fully configured and functional on the backend. Sessions, skills, memory, config, and jobs all load correctly, only MCP fails.

Root cause:
The workspace calls GET /api/mcp on the dashboard, which doesn't exist:
curl http://127.0.0.1:9119/api/mcp
→ {"detail":"No such API endpoint: /api/mcp"}
The correct endpoint is /api/mcp/servers, which works and returns real data:
curl http://127.0.0.1:9119/api/mcp/servers
→ {"servers":[{"name":"filesystem","transport":"stdio",...,"enabled":true}]}
Confirmation the backend itself is fine:

hermes mcp list inside the agent container shows the server correctly:

filesystem npx -y @modelcontextprotocol/server-github... all ✓ enabled

Asking the agent to use the tool in chat works end to end (e.g. "list the files in /tmp" returns real results via the MCP filesystem server)

Expected: the workspace's MCP page should call /api/mcp/servers (matching the dashboard's actual route) instead of /api/mcp.

Suggested fix: update whatever client/proxy code in hermes-workspace constructs the MCP status request to append /servers to the base MCP path.