Run: 2026-05-08 (scheduled, autonomous) · Source: ~/.claude.json user-scope mcpServers

Executive summary

Per-server table

Server Transport Status Tools Server info
callrail stdio connected 50 callrail-mcp v1.27.0
cloudflare-dns-analytics http oauth-pending n/a 401 invalid_token
cloudflare-docs http connected n/a docs-ai-search v0.4.7
cloudflare-observability http oauth-pending n/a 401 invalid_token
cloudways http connected n/a cloudways-mcp-server v1.27.0
google-ads stdio connected 3 Google Ads Server v3.2.3
ringcentral stdio connected 48 mcp-server-ringcentral v1.0.0
sentry http oauth-pending n/a 401 invalid_token

Tool counts for HTTP servers were not captured: tools/list requires session continuation after the initialize handshake, and the canonical task body explicitly accepts 200/401 from initialize as the connectivity signal. Stdio probes do the full initialize → tools/list round-trip.

Drift detail

None. First run → baseline established at snapshots/2026-05-08/ (8 files).

Action items

  1. OAuth pending — Sam to authorize when next interactive:
  2. Probe-script note (low priority): First-pass probe used Python urllib default UA, which Cloudflare-fronted endpoints fingerprinted and 403'd. Switched to Claude-Code-MCP-HealthCheck/1.0 UA + MCP-Protocol-Version header and all five HTTP endpoints responded normally. Future runs default to the second probe.
  3. Escalation watch: No server in degraded state today. Tomorrow's run compares against snapshots/2026-05-08/ and escalates any flip to timeout/spawn-failed/http-error (non-401).

Files written this run