otel-mcp-server
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| OPA_URL | No | Open Policy Agent REST API URL. Disabled when unset. | |
| LOKI_URL | No | Loki API URL. | http://localhost:3100 |
| ALLOY_URL | No | Grafana Alloy URL. Disabled when unset. | |
| BEATS_URL | No | Beats HTTP monitoring endpoint URL. Disabled when unset. | |
| TEMPO_URL | No | Grafana Tempo URL (used when TRACES_PROVIDER=tempo). | http://localhost:3200 |
| CILIUM_URL | No | Cilium agent REST API URL. Disabled when unset. | |
| CONSUL_URL | No | Consul HTTP API URL. Disabled when unset. | |
| INFLUX_URL | No | InfluxDB HTTP API URL. Disabled when unset. | |
| JAEGER_URL | No | Jaeger Query API URL (used when TRACES_PROVIDER=jaeger). | http://localhost:16686 |
| VECTOR_URL | No | Vector API URL. Disabled when unset. | |
| ZIPKIN_URL | No | Zipkin v2 API URL (used when TRACES_PROVIDER=zipkin). | http://localhost:9411 |
| APP_API_URL | No | Application API URL for ZK/system tools. | http://localhost:5000 |
| GRAFANA_URL | No | Grafana API URL. Disabled when unset. | |
| GRAYLOG_URL | No | Graylog REST API URL. Disabled when unset. | |
| TRAEFIK_URL | No | Traefik API URL. Disabled when unset. | |
| VMALERT_URL | No | vmalert rules and alerts API URL. Disabled when unset. | |
| MCP_BACKENDS | No | JSON array giving full control over multi-backend instances, including explicit product and per-instance headers. Takes precedence over env-var URL forms. | |
| MCP_JIT_MODE | No | JIT privileged identity mode: off, enabled, or required. | off |
| OPENTSDB_URL | No | OpenTSDB HTTP API URL. Disabled when unset. | |
| PINPOINT_URL | No | Pinpoint web API URL. Disabled when unset. | |
| FLUENTBIT_URL | No | Fluent Bit HTTP monitoring server URL. Disabled when unset. | |
| MCP_AUTH_KEYS | No | JSON string containing client API keys, best for containers or Kubernetes Secrets. | |
| MCP_JIT_STORE | No | Where JIT token and single-use replay state lives. memory is in-process; any other value is a module specifier for an operator-supplied shared-store adapter. | memory |
| PYROSCOPE_URL | No | Pyroscope HTTP API URL. Disabled when unset. | |
| AGENTRELAY_URL | No | AgentRelay hosted REST API URL for agent coordination. Disabled when unset. | |
| CLICKHOUSE_URL | No | ClickHouse HTTP interface URL. Disabled when unset. | |
| GRAFANA_ORG_ID | No | Sets the X-Grafana-Org-Id header for multi-org Grafana. | |
| KONG_ADMIN_URL | No | Kong Admin API URL. Disabled when unset. | |
| KUBERNETES_URL | No | kube-apiserver URL; auto-detected in-cluster via the ServiceAccount mount when unset. | |
| LOKI_TENANT_ID | No | Sets the X-Scope-OrgID header for multi-tenant Loki. | |
| MCP_TIMEOUT_MS | No | Backend query timeout in milliseconds. | 15000 |
| PROMETHEUS_URL | No | Prometheus API URL. | http://localhost:9090 |
| SKYWALKING_URL | No | SkyWalking OAP GraphQL URL (used when TRACES_PROVIDER=skywalking). | http://localhost:12800 |
| ENVOY_ADMIN_URL | No | Envoy admin API URL. Disabled when unset. | |
| TRACES_PROVIDER | No | Trace backend selector — one of jaeger, tempo, zipkin, skywalking. | jaeger |
| ALERTMANAGER_URL | No | Alertmanager API URL. Disabled when unset. | |
| KUBERNETES_TOKEN | No | ServiceAccount bearer token for Kubernetes when running out-of-cluster. | |
| TRACES_TEMPO_URL | No | Grafana Tempo URL alias (used when TRACES_PROVIDER=tempo). | http://localhost:3200 |
| CONSUL_AUTH_TOKEN | No | Consul ACL token. | |
| ELASTICSEARCH_URL | No | Elasticsearch / OpenSearch API URL. Disabled when unset. | |
| GRAFANA_MAX_ITEMS | No | Default Grafana list/search limit. | 50 |
| MCP_ENABLE_WRITES | No | Enable mutating/write tools. Read-only by default; accepts true, 1, yes, or on. | |
| TRACES_JAEGER_URL | No | Jaeger Query API URL alias (used when TRACES_PROVIDER=jaeger). | http://localhost:16686 |
| TRACES_ZIPKIN_URL | No | Zipkin v2 API URL alias (used when TRACES_PROVIDER=zipkin). | http://localhost:9411 |
| KUBERNETES_CA_FILE | No | Path to the Kubernetes cluster CA file used for TLS validation. | |
| MCP_AUTH_KEYS_FILE | No | Path to a JSON file containing client API keys, e.g. a mounted Kubernetes Secret. | |
| MCP_VERSION_GATING | No | Version-gating mode: off, warn, or enforce. | |
| MCP_JIT_TTL_SECONDS | No | JIT token TTL and cap for requested ttlSeconds, clamped to 60–3600. | 900 |
| MCP_SESSION_IDLE_MS | No | HTTP transport only: idle time before an inactive session is reaped, in milliseconds. | 300000 |
| BEYLA_PROMETHEUS_URL | No | Prometheus store holding Grafana Beyla eBPF metrics. Disabled when unset. | |
| GRAFANA_DEFAULT_FROM | No | Default Grafana query range start. | now-1h |
| MCP_SESSION_SWEEP_MS | No | HTTP transport only: how often the idle-session reaper runs, in milliseconds. | 60000 |
| AGENTRELAY_AUTH_TOKEN | No | AgentRelay bearer token. | |
| KUBERNETES_TOKEN_FILE | No | Path to a file containing the Kubernetes ServiceAccount bearer token. | |
| TRACES_SKYWALKING_URL | No | SkyWalking OAP GraphQL URL alias (used when TRACES_PROVIDER=skywalking). | http://localhost:12800 |
| AGENTRELAY_AUTH_HEADER | No | Raw header value for AgentRelay, e.g. for an X-API-Key scheme. | |
| PROMETHEUS_PATH_PREFIX | No | Path prefix for Prometheus, e.g. /prometheus. | |
| MCP_JIT_MAX_ACTIVE_TOKENS | No | Active JIT token capacity guard. | 1000 |
| MCP_ENTERPRISE_AUTH_ISSUER | No | Trusted enterprise IdP issuer URL. Required only when enabling enterprise-managed authorization. | |
| MCP_ENTERPRISE_AUTH_AUDIENCE | No | This server's issuer identifier; the ID-JAG aud must contain it. | |
| MCP_ENTERPRISE_AUTH_JWKS_URL | No | Explicit IdP JWKS endpoint; defaults to OIDC discovery from the issuer. | |
| MCP_ENTERPRISE_AUTH_RESOURCE | No | This server's resource identifier; the ID-JAG resource must match it. Defaults to the audience. | |
| MCP_JIT_MAX_LIFETIME_SECONDS | No | Hard cap per JIT token lineage, clamped to ttl–86400. | 28800 |
| MCP_ENTERPRISE_AUTH_DEFAULT_SCOPES | No | Scopes granted when an ID-JAG carries no scope claim; defaults to all enabled skills. | |
| MCP_ENTERPRISE_AUTH_MAX_REDEEMED_JTIS | No | Redeemed-jti replay-cache cap; fail-closed at the cap. | 50000 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
| resources | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| traces_searchB | Search distributed traces. Backend: Jaeger. Param semantics vary by provider: |
| trace_getB | Get full trace detail by ID. Backend: Jaeger. |
| traces_servicesC | List services known to the traces backend. Backend: Jaeger. |
| traces_operationsC | List operations/endpoints for a service. Backend: Jaeger. |
| traces_dependenciesC | Service dependency graph — which services call which. Backend: Jaeger. |
| metrics_queryB | Execute an instant PromQL query and return current metric values. |
| metrics_query_rangeC | Execute a range PromQL query and return a time series. |
| metrics_targetsA | List all Prometheus scrape targets and their health status. |
| metrics_alertsA | Get all Prometheus alerting rules and their current state (firing, pending, inactive). |
| metrics_metadataA | Look up metric metadata — type, help text, and unit for a metric name. |
| metrics_label_valuesB | Get all values for a given Prometheus label (e.g. list all jobs, instances, or status codes). |
| logs_queryB | Query logs from Loki using LogQL. Returns log lines matching the query. |
| logs_labelsB | List all label names available in Loki. |
| logs_label_valuesA | Get all values for a Loki label (e.g. all app names, namespaces, components). |
| logs_tail_contextA | Get recent logs correlated with a specific trace ID. Searches across all apps for log lines containing the trace ID. |
| zk_proof_getB | Retrieve a ZK-SNARK proof for a specific trade. Returns the Groth16 proof, public signals, and verification key. |
| zk_proof_verifyB | Verify a ZK-SNARK proof server-side. Returns whether the Groth16 proof is mathematically valid. |
| zk_solvencyB | Get the latest solvency proof — proves total reserves ≥ liabilities without revealing individual balances. |
| zk_statsA | Get aggregate ZK proof statistics — total proofs generated, verification success rate, average proving time, circuit breakdown. |
| anomalies_activeC | Get currently active anomalies detected by trace-based and amount-based detectors. |
| anomalies_baselinesA | Get current span duration baselines used for anomaly detection — mean, stdDev, p50, p95, p99 per operation. |
| system_healthA | Get full system health — service status, uptime, performance metrics, active alerts. |
| system_topologyB | Get the live service dependency topology with health overlays from Jaeger and the application API. |
| backend_capabilitiesB | Report supported product versions and protocol feature availability per backend. Optionally filter by skill and classify a specific backend version into its support tier with a per-feature availability breakdown. |
| exchange_statusB | Get the KrystalineX exchange overall status: operational state, uptime, current observability posture, and last incident summary. |
| total_volumeA | Get aggregate trading volume on the KrystalineX exchange (24h, weekly, all-time). Returns volume per asset and totals. |
| recent_tradesB | Get the most recent anonymized public trades on KrystalineX. Useful for "is the market active?" questions. Trades carry no user identity. |
| transparency_metricsB | Get the full transparency metrics bundle: trace coverage, anomaly counts, MTTR, ZK proof generation rate, etc. Powers the public "Proof of Observability" dashboard. |
| verify_traceA | Look up the public-safe distributed trace for a given trade ID — service hops, timing, and verification result. Useful for "show me what happened to this trade". |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| platform-overview |
TDQS
Scored across 29 tools
Most tools target a distinct backend+resource (metrics_query vs logs_query vs traces_search), and domain prefixes make routing obvious. There is mild overlap among the summary-style tools — system_health, exchange_status, and transparency_metrics all surface posture/alerts/uptime, and verify_trace reads like a verification verb but is actually a business-level trace lookup (contrast zk_proof_verify). These edge cases are clarified by descriptions, so misselection risk is low but not zero.
The dominant pattern is snake_case with a domain prefix (metrics_*, traces_*, logs_*, zk_*, anomalies_*, system_*), which is readable and mostly predictable. However, six tools drop the prefix entirely (exchange_status, total_volume, recent_trades, transparency_metrics, verify_trace, backend_capabilities), and suffix styles mix verbs and nouns (traces_services vs traces_search, zk_solvency vs zk_proof_get), so the convention is not fully uniform.
At 29 tools the surface is on the heavy side and exceeds the comfortable 3-15 band, with each domain contributing 4-6 tools. The breadth is justified by the server spanning four distinct backends (Prometheus, Jaeger, Loki, ZK) plus a business-exchange domain, so it is borderline rather than bloated. A few summary tools (exchange_status, transparency_metrics, system_health) could be consolidated.
Coverage is strong across every advertised domain: metrics (targets, instant/range queries, alerts, metadata, label values), traces (get, services, operations, dependencies, search), logs (query, label names/values, trace correlation), ZK (proof get/verify, solvency, stats), anomalies, and system topology/health. Minor gaps remain — there is no Prometheus label-name listing to mirror logs_labels, no span-level retrieval, and no ZK proof history/listing — but agents can work around these via the query tools.