MCP Server Tools

The Kloudfuse MCP server exposes a set of tools that an AI assistant calls to read your observability data. Every tool is read-only: it queries metrics, logs, traces, alerts, dashboards, Kubernetes state, profiles, RUM, or documentation, and never changes anything.

Tools are grouped into toolsets. On the remote MCP server, you can limit which toolsets a connection can see by appending ?toolsets=<names> to the MCP server URL as described in Toolsets. The local Docker server does not read this query parameter: it always exposes every default toolset, and the opt-in LogQL toolset cannot be enabled there.

In This Topic

Metrics

Run PromQL queries against your metrics and explore label names and values.

query_prometheus

Toolset: metrics

Runs a PromQL expression against your metrics and returns an instant or range result.

  • Show me CPU usage per pod in the kfuse namespace over the last hour.

  • What is the p99 request latency for the checkout service right now?

  • Plot memory usage for the ingester over the past 6 hours.

search_prometheus_label_names

Toolset: metrics

Lists the label names available on your metrics, optionally scoped to a series selector.

  • What labels are available on container_cpu_usage_seconds_total?

  • List all labels in the metrics store.

search_prometheus_label_values

Toolset: metrics

Lists the unique values of one metric label.

  • Which namespaces are reporting metrics?

  • List every value of the kube_cluster_name label.

Logs

Query logs with FuseQL, read the lines around a match, group logs by fingerprint, and discover labels and facets.

query_logs

Toolset: logs

Queries logs with FuseQL and returns either raw log lines or an aggregated metric table.

  • Show me error logs from the payments service in the last 30 minutes.

  • Count log lines by level for the api service over the past hour.

  • Find logs containing "connection refused" in the production namespace.

get_log_context

Toolset: logs

Fetches the raw log lines immediately before and after one log line, like scrolling around it in the log viewer.

  • Show me what happened right before this out-of-memory log line.

  • Get 50 lines of context around that stack trace.

search_log_fingerprints

Toolset: logs

Groups logs into fingerprint patterns with occurrence counts to surface the noisiest and most common log templates.

  • What are the most common log patterns from the gateway in the last hour?

  • Which services are producing the most repeated error messages?

get_log_label_names

Toolset: logs, logql

Lists the label names available on logs.

  • What labels can I filter logs by?

get_log_label_values

Toolset: logs, logql

Lists the values of one log label.

  • Which services are sending logs?

  • List all Kubernetes namespaces present in logs.

get_log_facet_names

Toolset: logs, logql

Lists the facet names available on logs, with their data types.

  • What parsed fields are available on nginx logs?

get_log_facet_values

Toolset: logs, logql

Lists the values of one log facet, with counts.

  • What HTTP status codes appear in the logs, and how often?

  • Which hosts show up in the hostname facet?

APM

Inspect services and dependencies, search traces, and drill into waterfalls, spans, and errors.

search_apm_services

Toolset: apm

Finds active APM services and databases in your infrastructure.

  • Which services are running in production?

  • List all databases the APM data knows about.

search_apm_dependencies

Toolset: apm

Returns service-to-service dependencies to build a service graph.

  • What does the checkout service depend on?

  • Which services call the postgres database?

get_apm_service_nodes

Toolset: apm

Lists the Kubernetes nodes that handled traffic for a service.

  • Which nodes served traffic for the frontend service today?

search_apm_traces

Toolset: apm

Searches traces by service, operation, duration, status, and other criteria.

  • Find traces slower than 2 seconds for the checkout service in the last hour.

  • Show me failed traces for the POST /orders endpoint.

  • List the slowest traces in production from the past 15 minutes.

get_apm_trace_waterfall

Toolset: apm

Returns the full span waterfall for one trace ID.

  • Show me the waterfall for trace abc123.

  • Where is the time going in this trace?

get_apm_span_details

Toolset: apm

Returns complete details for one span, given its span ID and trace ID.

  • Show me the attributes of the database span in trace abc123.

  • What tags are set on this span?

search_apm_trace_errors

Toolset: apm

Searches trace errors matching the given criteria.

  • What errors is the payments service throwing right now?

  • Group recent trace errors by exception type.

get_apm_label_names

Toolset: apm

Lists the label names available on traces and spans.

  • What attributes can I filter traces by?

get_apm_label_values

Toolset: apm

Lists the values of one trace or span label.

  • Which operations does the checkout service expose?

  • List all values of http.status_code seen in traces.

Alerts

Search alert rules and their firing instances, read the queries behind a rule, and review alert history.

get_alert_summary

Toolset: alerts

Returns a high-level rollup of the cluster’s current alert state: what is firing and what has been changing.

  • What is firing right now?

  • Give me an overview of alert activity today.

search_alert_rules

Toolset: alerts

Searches alert rules by name, labels, severity, health, or state, and returns each rule’s current state.

  • List all critical alert rules.

  • Which alert rules mention Kafka?

  • Are any alert rules in an error state?

search_alert_instances

Toolset: alerts

Lists the individual firing instances of a rule, one per evaluated label set, so you can see which pods or services are affected.

  • Which pods are firing the high memory alert?

  • Show me every firing instance in the kfuse namespace.

get_alert_rule_queries

Toolset: alerts

Returns the full query and detection pipeline behind one alert rule, for any signal type and detection method.

  • What query does the disk usage alert evaluate?

  • Show me the threshold and condition for this alert rule.

get_alert_history

Toolset: alerts

Returns historical state transitions for alert rules, either as raw rows or an aggregated rollup, to spot past fires and flapping.

  • When did the latency alert fire in the last 7 days?

  • Which alerts have been flapping this week?

  • How long was the Kafka lag alert firing yesterday?

Dashboards

Search dashboards and folders, list tags, and read dashboard definitions and panel queries.

search_dashboards

Toolset: dashboards

Searches dashboards by title, tag, folder, or starred status.

  • Find dashboards related to Kafka.

  • List my starred dashboards.

search_folders

Toolset: dashboards

Searches dashboard folders.

  • What dashboard folders exist?

  • Find the folder for the platform team.

list_dashboard_tags

Toolset: dashboards

Lists dashboard tags with the number of dashboards using each.

  • Which dashboard tags are in use?

get_dashboard

Toolset: dashboards

Returns a dashboard definition by UID.

  • Show me the definition of the Pinot overview dashboard.

  • What panels are on dashboard abc123?

get_dashboard_panel_queries

Toolset: dashboards

Returns the query expressions behind a dashboard’s panels, grouped by row, optionally filtered by a title pattern or panel ID.

  • What query powers the "Consumer lag" panel?

  • Show me all the queries in the Ingestion section of the Kafka dashboard.

Events

Query events with filtering, sorting, and facet discovery.

get_events

Toolset: events

Returns events in a time range with filtering, sorting, and pagination.

  • Show me Kubernetes events from the last hour.

  • Find pod restart events in the kfuse namespace today.

  • List warning events for the ingester deployment.

get_event_label_names

Toolset: events

Lists the label names available on events in a time range.

  • What fields can I filter events by?

get_event_facet_values

Toolset: events

Lists the values of one event facet, with counts.

  • What event reasons occurred most in the last 24 hours?

  • Which namespaces generated events today?

Infrastructure

Search Kubernetes objects and read their status.

search_kubernetes

Toolset: infrastructure

Queries Kubernetes objects such as pods, deployments, services, and nodes, returning only the fields you request.

  • List pods that are not running in the kfuse namespace.

  • How many replicas does the ingester deployment have?

  • Which nodes are in the cluster and what are their capacities?

get_kubernetes_object_status

Toolset: infrastructure

Returns the status section of Kubernetes objects, including conditions and phase history.

  • Why is the pinot-server statefulset not ready?

  • Show me the conditions on node worker-3.

Pyroscope

Pull continuous-profiling data: profile types, labels, series, flamegraphs, and diffs between time windows.

get_pyroscope_profile_types

Toolset: pyroscope

Lists the available continuous-profiling profile types.

  • What profile types are being collected?

get_pyroscope_label_names

Toolset: pyroscope

Lists the label names available for filtering profiles.

  • What labels can I filter profiles by?

get_pyroscope_label_values

Toolset: pyroscope

Lists the values of one profile label.

  • Which services have profiling data?

get_pyroscope_series

Toolset: pyroscope

Finds the unique series, or label combinations, that match a selector.

  • What profile series exist for the query service?

get_pyroscope_flamegraph

Toolset: pyroscope

Returns flamegraph data for a profile type and label selector.

  • Show me the CPU flamegraph for the ingester over the last 15 minutes.

  • Where is the query service spending its time?

get_pyroscope_time_series

Toolset: pyroscope

Returns profile values over time as a time series.

  • Plot memory allocation for the ingester over the last hour.

diff_pyroscope_profiles

Toolset: pyroscope

Compares profiles between two time windows to find regressions.

  • Compare CPU profiles for the query service before and after the deploy at 2pm.

  • What functions got slower in the last hour compared to the hour before?

RUM

Real User Monitoring: list applications, search events, view error groups, latency, time series, and unminified stack traces.

list_rum_applications

Toolset: rum

Lists all RUM applications, or returns details for one.

  • What RUM applications are configured?

  • Show me details for the web storefront application.

search_rum_events

Toolset: rum

Searches RUM events within one application.

  • Show me error events from the storefront app in the last hour.

  • Find page views for the checkout page from Safari users.

get_rum_attributes

Toolset: rum

Lists the available RUM attribute fields, organized by category and event type.

  • What attributes can I filter RUM events by?

get_rum_attribute_values

Toolset: rum

Lists the values of one RUM attribute within an application, with counts.

  • Which browsers are users on?

  • What are the most visited pages today?

get_rum_time_series

Toolset: rum

Returns a time series for one field and aggregation within an application.

  • Plot the error rate for the storefront app over the last 24 hours.

  • Show me page view counts per hour today.

get_rum_raw_event

Toolset: rum

Returns the complete raw data for one RUM event, by event ID and type.

  • Show me the full payload for event abc123.

get_rum_error_groups

Toolset: rum

Returns error groups with occurrence counts within an application.

  • What are the top frontend errors in the last 24 hours?

  • Which JavaScript errors are new since yesterday?

get_rum_latency_distribution

Toolset: rum

Returns a latency histogram for one field within an application.

  • What does the page load time distribution look like this week?

  • Show me the distribution of Largest Contentful Paint for the checkout page.

unminify_rum_stacktrace

Toolset: rum

Unminifies a frontend error stack trace to show original source files, line numbers, and surrounding code.

  • Unminify this stack trace and tell me where the error is thrown.

  • Which source file does this minified error come from?

Docs

Search and read the Kloudfuse documentation.

search_kloudfuse_docs

Toolset: docs

Searches the Kloudfuse documentation for pages matching a natural-language or keyword query.

  • How do I write a FuseQL aggregation?

  • Find the docs on configuring alert contact points.

get_kloudfuse_doc

Toolset: docs

Returns the full content of one Kloudfuse documentation page.

  • Read the full FuseQL parse operators page.

  • Show me the page on MCP toolsets.

LogQL

Legacy LogQL tools. This toolset is opt-in only: its tools are hidden unless you request the toolset explicitly with ?toolsets=logql. The log label and facet tools listed under Logs are also included in this toolset.

search_logs

Toolset: logql

Queries application logs with LogQL filters.

  • Show me error logs from the payments service in the last 30 minutes.

  • Find logs containing "timeout" for the api pod.

query_log_metrics

Toolset: logql

Calculates log-based metrics over time with a LogQL expression.

  • What is the error log rate for the gateway over the last hour?

  • Count log lines per service in the last 15 minutes.