Operations

This section is the production hardening reference for operators running discord-mcp at scale. The other sections — Tools, Recipes, and Architecture — describe what the server can do and how it’s built. This section is about how to run it safely and observably in front of a real Discord guild and a real agent.

Every page here is rooted in the env-var contract: the server’s runtime behavior is fully controlled by environment variables (no config files, no mutable runtime state). Below are the four operator concerns, each with its own page describing defaults, override knobs, and the observability surface.

Telemetry Enable OpenTelemetry traces and metrics. Honeycomb, local Jaeger, console exporter; six built-in metrics; trace context propagation into Discord REST.

Resilience The composite policy chain — bulkhead, circuit breaker, retry, timeout. 11 environment variables; 429-aware backoff; pipeline interactions.

Audit logging Mutating-only audit trail. Four sinks (stderr, file, OTLP-stub, none); JSONL schema; PII redaction policy; SOC2/HIPAA considerations.

Client capability matrix Which MCP clients support which capabilities (tools, resources, subscribe, sampling, elicitation). Setup links per client.

Quick environment-variable orientation

If you only read one section, read Resilience — it’s the one set of knobs that affects every Discord call the server makes. Telemetry is opt-in (OTEL_ENABLED=false by default, zero overhead) and audit is on by default but can be tuned per-sink.

For per-tool environment variables (token, scopes, allowed guilds, dry-run mode), see the Reference → Environment variables section once Phase E lands.

Looking for something else?

How to set up your client → Get started → Client setup.
Single-tool reference → Tools.
Multi-step workflows → Recipes.
How the server is built → Architecture.