Grafana MCP server

A [Model Context Protocol][mcp] (MCP) server for Grafana.

This provides access to your Grafana instance and the surrounding ecosystem.

Quick Start

Requires uv. Add the following to your MCP client configuration (e.g. Claude Desktop, Cursor):

{
  "mcpServers": {
    "grafana": {
      "command": "uvx",
      "args": ["mcp-grafana"],
      "env": {
        "GRAFANA_URL": "http://localhost:3000",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<your service account token>"
      }
    }
  }
}

For Grafana Cloud, replace GRAFANA_URL with your instance URL (e.g. https://myinstance.grafana.net). See Usage for more installation options including Docker, binary, and Helm.

Requirements

• Grafana version 9.0 or later is required for full functionality. Some features, particularly datasource-related operations, may not work correctly with earlier versions due to missing API endpoints.

Features

The following features are currently available in MCP server. This list is for informational purposes only and does not represent a roadmap or commitment to future features.

Dashboards

• Search for dashboards: Find dashboards by title, folder UID, tag, or starred status
• Get dashboard by UID: Retrieve full dashboard details using its unique identifier. Pass optional version to load a saved snapshot instead of the current dashboard. Warning: Large dashboards can consume significant context window space.
• List dashboard versions: List saved versions of a dashboard as compact metadata (version number, author, timestamp, save message)
• Get dashboard summary: Get a compact overview of a dashboard including title, panel count, panel types, variables, and metadata without the full JSON to minimize context window usage
• Get dashboard property: Extract specific parts of a dashboard using JSONPath expressions (e.g., $.title, $.panels[*].title) to fetch only needed data and reduce context window consumption
• Update or create a dashboard: Modify existing dashboards or create new ones. Warning: Requires full dashboard JSON which can consume large amounts of context window space.
• Patch dashboard: Apply specific changes to a dashboard without requiring the full JSON, significantly reducing context window usage for targeted modifications
• Get panel queries and datasource info: Get the title, query string, and datasource information (including UID and type, if available) from every panel in a dashboard

Run Panel Query

Note: Run panel query tools are disabled by default. To enable them, add runpanelquery to your --enabled-tools flag.

• Run panel query: Execute a dashboard panel's query with custom time ranges and variable overrides.

Context Window Management

The dashboard tools now include several strategies to manage context window usage effectively (issue #101):

• Use get_dashboard_summary for dashboard overview and planning modifications
• Use get_dashboard_property with JSONPath when you only need specific dashboard parts
• Avoid get_dashboard_by_uid unless you specifically need the complete dashboard JSON

Datasources

• List and fetch datasource information: View all configured datasources and retrieve detailed information about each.
  • Supported datasource types: Prometheus, Loki, ClickHouse, CloudWatch, Elasticsearch, OpenSearch, Snowflake, Athena.

Query Examples

Note: Query examples tools are disabled by default. To enable them, add examples to your --enabled-tools flag.

• Get query examples: Retrieve example queries for different datasource types to learn query syntax.

Prometheus Querying

• Query Prometheus: Execute PromQL queries (supports both instant and range metric queries) against Prometheus datasources.
• Query Prometheus metadata: Retrieve metric metadata, metric names, label names, and label values from Prometheus datasources.
• Query histogram percentiles: Calculate histogram percentile values (p50, p90, p95, p99) using histogram_quantile.

Loki Querying

• Query Loki logs and metrics: Run both log queries and metric queries using LogQL against Loki datasources.
• Query Loki metadata: Retrieve label names, label values, and stream statistics from Loki datasources.
• Query Loki patterns: Retrieve log patterns detected by Loki to identify common log structures and anomalies.

InfluxDB Querying

Note: InfluxDB tools are disabled by default. To enable them, add influxdb to your --enabled-tools flag.

• Query InfluxDB: Execute queries against InfluxDB datasources using either InfluxQL (v1.x) or Flux (v2.x). The dialect is inferred from the datasource configuration, or can be set explicitly via the dialect parameter.

SQL Datasource Querying

Note: SQL tools are disabled by default. To enable them, add sql to your --enabled-tools flag. Back-compat aliases clickhouse, snowflake, and athena also work.

Unified SQL tools support ClickHouse, Snowflake, Athena, MySQL, PostgreSQL, and MSSQL through a single set of tools. Queries go through Grafana's datasource plugins, so authentication is handled by datasource configuration — credentials are never seen by the MCP server.

• List databases/schemas/catalogs: Discover organizational units for a SQL datasource. For Athena, omit catalog to list catalogs, or pass a catalog to list databases.
• List tables: List tables in a database or schema with metadata (row counts, sizes where available).
• Describe table schema: Get column names, types, nullability, defaults, and comments.
• Query SQL: Execute SQL queries with datasource-specific macro substitution ($__timeFilter(col), $__from/$__to, $__interval, ${varname}), automatic limit enforcement, and template variable support.

CloudWatch Querying

Note: CloudWatch tools are disabled by default. To enable them, add cloudwatch to your --enabled-tools flag.

• List CloudWatch namespaces: Discover available AWS CloudWatch namespaces.
• List CloudWatch metrics: List metrics available in a specific namespace.
• List CloudWatch dimensions: Get dimensions for filtering metric queries.
• Query CloudWatch: Execute CloudWatch metric queries with time range support.

Graphite Querying

Note: Graphite tools are disabled by default. To enable them, add graphite to your --enabled-tools flag.

• Query Graphite: Execute Graphite render API queries against a Graphite datasource.
• List Graphite metrics: Browse and discover Graphite metric paths.
• List Graphite tags: List available Graphite tags and tag values.
• Query Graphite density: Query Graphite metric density for a given pattern.

Elasticsearch/OpenSearch Querying

Note: Elasticsearch/OpenSearch tools are disabled by default. To enable them, add elasticsearch to your --enabled-tools flag.

• Query Elasticsearch/OpenSearch: Execute search queries against Elasticsearch or OpenSearch datasources using either Lucene query syntax or Elasticsearch Query DSL. Supports filtering by time range and retrieving logs, metrics, or any indexed data. Returns documents with their index, ID, source fields, and optional relevance score.

Quickwit Querying

Note: Quickwit tools are disabled by default. To enable them, add quickwit to your --enabled-tools flag.

• Query Quickwit: Execute search queries against Quickwit datasources using Lucene query syntax or partial Elasticsearch-compatible Query DSL. Supports filtering by time range and retrieving logs or other indexed documents. Returns documents with their index, ID, source fields, and optional relevance score.

Agent Observability

Note: Agent Observability tools are disabled by default and work only in Grafana Cloud. To enable them, add agento11y to your --enabled-tools flag.

• List and search conversations: List recent LLM conversations or search them with a filter expression (model, provider, agent, status, error type, eval results, and more) over a time range. Search results include error counts, rating summaries, evaluation summaries, and trace IDs.
• Get conversation detail: Fetch a single conversation with all its generations, including prompts and outputs.
• Get generation detail and scores: Fetch a single generation by ID, and its evaluation scores (evaluator, score key, value, passed, explanation).
• Read the agent catalog: List the agents that send telemetry, fetch one agent version in full (complete system prompt, every tool with its JSON schema, and the models it ran on), walk an agent's version history, and compare evaluation score aggregates per version. Effective versions are sha256: hashes that a tool change never affects; for an agent that reports no version of its own they hash the system prompt, so a prompt edit mints a new version. Catalog and version rows carry a token_estimate, which is worth checking before fetching a full prompt.
• Inspect evaluators and templates: Read the evaluators a score came from, the templates they were derived from, and the judge providers and models available to LLM-judge evaluators. With write tools enabled, also create, fork, test, and delete evaluators.
• Inspect eval rules and guards: Read the asynchronous eval rules that bind evaluators to production traffic, and the guards (hook rules) that run inline and can warn or deny. With write tools enabled, also create, update, preview, and delete them. Writes and the non-persisting preview_rule and test_evaluator operations need the grafana-agento11y-app.eval:write permission, granted by the Agento11y Admin role.
• Curate saved conversations and collections: Read the saved conversations (bookmarks that give a conversation a stable ID, name, and tags) and the collections that group them, including each collection's member count and the collections embedded in every saved-conversation row. With write tools enabled, also bookmark a conversation, create and edit collections, and add or remove members. These writes need the same grafana-agento11y-app.eval:write permission.
• Read and edit test suites: List the versioned test suites that offline experiments run against, read one with its full version history, and page through the test cases of a version. With write tools enabled, also create a suite, rename or retag it, open a draft version, publish it, and write or delete its test cases. A published version is frozen, so an edit means opening a new draft. These writes need grafana-agento11y-app.eval:write.
• Read offline experiments: List the evaluation runs over a test suite and read one with its headline pass rate, cost, and token totals. Drill down through a per-test-case report to trials, their scores with each judge's explanation, and their artifact metadata. With write tools enabled, also rename or retag an experiment and cancel a running one, which need grafana-agento11y-app.eval:write. Experiments are created by SDK runners, not by this tool.

Grafana Assistant

Note: Assistant tools are disabled by default and require the Grafana Assistant plugin (grafana-assistant-app) to be installed on the target Grafana instance. They are also write tools (the assistant may mutate stack state), so they are skipped when --disable-write is set. To enable them, add assistant to your --enabled-tools flag.

• Ask the assistant: Send a natural-language prompt to Grafana Assistant and wait for the full text reply. The assistant may use tools, metrics, logs, and other stack context—broader than firing one isolated data-source query. Pass the returned contextId back in a follow-up call to continue the same conversation. Complex tasks can take several minutes; the call blocks until the reply is done or the request times out (5 minutes).

Incidents

• Search, create, and update incidents: Manage incidents in Grafana Incident, including searching, creating, adding activities, and reading or setting custom fields.

Sift Investigations

• List Sift investigations: Retrieve a list of Sift investigations, with support for a limit parameter.
• Get Sift investigation: Retrieve details of a specific Sift investigation by its UUID.
• Get Sift analyses: Retrieve a specific analysis from a Sift investigation.
• Find error patterns in logs: Detect elevated error patterns in Loki logs using Sift.
• Find slow requests: Detect slow requests using Sift (Tempo).

Alerting

• List and fetch alert rule information: View alert rules and their statuses (firing/normal/error/etc.) in Grafana. Supports both Grafana-managed rules and datasource-managed rules from Prometheus or Loki datasources.
• Create and update alert rules: Create new alert rules or modify existing ones.
• Delete alert rules: Remove alert rules by UID.
• Manage alerting routing: View notification policies, contact points, and time intervals. Supports both Grafana-managed contact points and receivers from external Alertmanager datasources (Prometheus Alertmanager, Mimir, Cortex).

Grafana OnCall

• List and manage schedules: View and manage on-call schedules in Grafana OnCall.
• Get shift details: Retrieve detailed information about specific on-call shifts.
• Get current on-call users: See which users are currently on call for a schedule.
• List teams and users: View all OnCall teams and users.
• List alert groups: View and filter alert groups from Grafana OnCall by various criteria including state, integration, labels, and time range.
• Get alert group details: Retrieve detailed information about a specific alert group by its ID.

Admin

Note: Admin tools are disabled by default. To enable them, include admin in your --enabled-tools flag.

• List teams: View all configured teams in Grafana.
• List Users: View all users in an organization in Grafana.
• List all roles: List all Grafana roles, with an optional filter for delegatable roles.
• Get role details: Get details for a specific Grafana role by UID.
• List assignments for a role: List all users, teams, and service accounts assigned to a role.
• List roles for users: List all roles assigned to one or more users.
• List roles for teams: List all roles assigned to one or more teams.
• List permissions for a resource: List all permissions defined for a specific resource (dashboard, datasource, folder, etc.).
• Describe a Grafana resource: List available permissions and assignment capabilities for a resource type.

User

• User info: Get the current Grafana identity — login, email, name, whether it is a Grafana (server) admin, the current organization, and the organizations the credential can access (with roles). Use it to discover valid orgId values for multi-organization requests.

Navigation

• Generate deeplinks: Create accurate deeplink URLs for Grafana resources instead of relying on LLM URL guessing.
  • Dashboard links: Generate direct links to dashboards using their UID (e.g., http://localhost:3000/d/dashboard-uid)
  • Panel links: Create links to specific panels within dashboards with viewPanel parameter (e.g., http://localhost:3000/d/dashboard-uid?viewPanel=5)
  • Explore links: Generate links to Grafana Explore with pre-configured datasources (e.g., http://localhost:3000/explore?schemaVersion=1&panes={"a":{"datasource":"prometheus-uid"}}). Grafana below 10.2 does not understand panes, so the legacy ?left={...} format is emitted for those versions instead.
  • Time range support: Add time range parameters to links (from=now-1h&to=now)
  • Custom parameters: Include additional query parameters like dashboard variables or refresh intervals

Annotations

• Get Annotations: Query annotations with filters. Supports time range, dashboard UID, tags, and match mode.
• Create Annotation: Create a new annotation on a dashboard or panel.
• Create Graphite Annotation: Create annotations using Graphite format (what, when, tags, data).
• Update Annotation: Replace all fields of an existing annotation (full update).
• Patch Annotation: Update only specific fields of an annotation (partial update).
• Delete Annotation: Permanently delete an annotation by ID.
• Get Annotation Tags: List available annotation tags with optional filtering.

Snapshots

• List snapshots: List dashboard snapshots with optional query and limit filters.
• Get snapshot: Retrieve snapshot metadata and dashboard payload by snapshot key.
• Create snapshot: Create a dashboard snapshot from a full dashboard payload, with optional expiration and external snapshot options.
• Delete snapshot: Delete a snapshot by snapshot key.

Rendering

• Get panel or dashboard image: Render a Grafana dashboard panel or full dashboard as a PNG image. Returns the image as base64 encoded data for use in reports, alerts, or presentations. Supports customizing dimensions, time range, theme, scale, and dashboard variables. Also supports rendering not-yet-applied dashboards from a provisioning repository branch (e.g. a git-sync PR preview) via the optional provisioningPreview parameter.
  • Note: Requires the Grafana Image Renderer service to be installed and configured.

Provisioning

• List provisioning repositories: List provisioning repositories configured for this Grafana instance (e.g. git-sync sources), returning each repository's slug along with its source URL, branch, path, sync state, and health.
• Validate provisioning file: Dry-run-apply a file from a provisioning repository at a given branch or commit. Returns whether it would be accepted, the resource action (create/update), the target resource type, and any structured validation errors — the same admission surface Grafana's PR commenter uses.

The list of tools is configurable, so you can choose which tools you want to make available to the MCP client. This is useful if you don't use certain functionality or if you don't want to take up too much of the context window. To disable a category of tools, use the --disable-<category> flag when starting the server. For example, to disable the OnCall tools, use --disable-oncall, or to disable navigation deeplink generation, use --disable-navigation.

RBAC Permissions

Each tool requires specific RBAC permissions to function properly. When creating a service account for the MCP server, ensure it has the necessary permissions based on which tools you plan to use. The permissions listed are the minimum required actions - you may also need appropriate scopes (e.g., datasources:*, dashboards:*, folders:*) depending on your use case.

Tip: If you're not familiar with Grafana RBAC or you want a quicker, simpler setup instead of configuring many granular scopes, you can assign a built-in role such as Editor to the service account. The Editor role grants broad read/write access that will allow most MCP server operations; it is less granular (and therefore less restrictive) than manually-applied scopes, so use it only when convenience is more important than strict least-privilege access.

Note: Grafana Incident and Sift tools use basic Grafana roles instead of fine-grained RBAC permissions:

• Viewer role: Required for read-only operations (list incidents, get investigations)
• Editor role: Required for write operations (create incidents, modify investigations)

For more information about Grafana RBAC, see the official documentation.

RBAC Scopes

Scopes define the specific resources that permissions apply to. Each action requires both the appropriate permission and scope combination.

Common Scope Patterns:

• Broad access: Use * wildcards for organization-wide access

  • datasources:* - Access to all datasources
  • dashboards:* - Access to all dashboards
  • folders:* - Access to all folders
  • teams:* - Access to all teams

• Limited access: Use specific UIDs or IDs to restrict access to individual resources

  • datasources:uid:prometheus-uid - Access only to a specific Prometheus datasource
  • dashboards:uid:abc123 - Access only to dashboard with UID abc123
  • folders:uid:xyz789 - Access only to folder with UID xyz789
  • teams:id:5 - Access only to team with ID 5
  • global.users:id:123 - Access only to user with ID 123

Examples:

• Full MCP server access: Grant broad permissions for all tools

  code

  datasources:* (datasources:read, datasources:query)
  dashboards:* (dashboards:read, dashboards:create, dashboards:write)
  folders:* (for dashboard creation and alert rules)
  teams:* (teams:read)
  global.users:* (users:read)
  

• Limited datasource access: Only query specific Prometheus and Loki instances

  code

  datasources:uid:prometheus-prod (datasources:query)
  datasources:uid:loki-prod (datasources:query)
  

• Dashboard-specific access: Read only specific dashboards

  code

  dashboards:uid:monitoring-dashboard (dashboards:read)
  dashboards:uid:alerts-dashboard (dashboards:read)
  

Tools

* Disabled by default. Add category to --enabled-tools to enable.

CLI Flags Reference

The mcp-grafana binary supports various command-line flags for configuration:

Transport Options:

• -t, --transport: Transport type (stdio, sse, or streamable-http) - default: stdio
• --address: The host and port for SSE/streamable-http server - default: localhost:8000
• --base-path: Base path for the SSE/streamable-http server
• --endpoint-path: Endpoint path for the streamable-http server - default: /mcp
• --server-name: Server name used in the MCP handshake and OTel service.name - default: mcp-grafana. Overrides GRAFANA_MCP_SERVER_NAME env var
• --instructions-append: Text appended to the server instructions returned to MCP clients on initialize, so every connecting agent sees it

HTTP Transport Security (SSE / streamable-http only):

Host/Origin validation is enforced on every route on the MCP listener — /sse, /mcp, and /healthz / /metrics when they share that listener — so a DNS-rebinding browser cannot reach any of them. Stdio transport is unaffected. --healthz-address and --metrics-address start a separate listener that is not wrapped.

• --allowed-hosts: Comma-separated allowlist of Host header values. Defaults to loopback variants of --address (e.g. localhost:8000,127.0.0.1:8000,[::1]:8000). A value that parses to empty (unset, ,, ,, etc.) also falls back to the defaults so a typo cannot silently disable the check. Requests with a Host header outside the allowlist are rejected with 403. Pass * to disable Host validation — only safe when a trusted reverse proxy validates Host. K8s httpGet probes and external /metrics scrapes will need either an explicit hostname in this list, *, a tcpSocket probe, or a separate port (--healthz-address / --metrics-address).
• --allowed-origins: Comma-separated allowlist of Origin header values. Empty by default — any request that carries an Origin header is rejected (browsers always send one for cross-origin requests, and no browser should be calling this server directly). Set to an explicit list to permit browser-based clients, or * to disable the check.

Caller Authentication (SSE / streamable-http only):

Optionally require MCP clients to authenticate to the server. This is separate from the credentials the server uses to reach Grafana. Stdio is unaffected.

• --server-auth-token: Bearer token callers must send as Authorization: Bearer <token>. Falls back to the MCP_GRAFANA_SERVER_TOKEN environment variable. When set, requests without a valid token are rejected with 401 before any tool runs. Prefer the env var so the secret isn't visible in the process arguments.

Caller authentication is enforced only when --server-auth-token is set. When it isn't and the server binds a non-loopback address, the server starts but logs a security error — emitted at the error log level so it isn't hidden by --log-level (loopback and stdio are unaffected); a future major release will make that a startup error. Use TLS (or TLS termination) whenever caller auth is enabled on a non-loopback address. When caller auth is enabled, the validated Authorization header is stripped before requests reach Grafana; combining --server-auth-token with GRAFANA_FORWARD_HEADERS=Authorization is rejected at startup.

Debug and Logging:

• --debug: Enable debug mode for detailed HTTP request/response logging
• --log-level: Log level (debug, info, warn, error) - default: info

Grafana Client Options:

• --grafana-timeout: Time limit for requests made by the Grafana client. Accepts Go duration strings (e.g., 10s, 500ms) - default: 10s
• --include-args-in-spans: Include tool call arguments in OpenTelemetry spans. Only enable in non-production environments or when arguments are known not to contain PII - default: false

Observability:

• --metrics: Enable Prometheus metrics endpoint at /metrics
• --metrics-address: Separate address for metrics server (e.g., :9090). If empty, metrics are served on the main server
• --healthz-address: Separate address for /healthz (e.g., :8080). If empty, /healthz is served on the main server. Shares a listener with --metrics-address when the two addresses match. Side listeners skip Host/Origin validation.
• --slow-request-threshold: Log an event when any MCP request (tool invocation, list, resource read, etc.) takes longer than this duration. Accepts Go duration strings (e.g., 500ms, 5s). Default 0 disables slow-request logging. See the Slow-request logging section.
• --slow-request-log-level: Log level for slow-request events (info or warn) - default: warn.

Anonymous Usage Statistics:

• --usage-stats: Anonymous usage statistics reporting: enabled, disabled, or log (print the report that would be sent to stderr and send nothing). Overrides the GRAFANA_USAGE_STATS env var, which in turn overrides DO_NOT_TRACK; any unrecognised value disables reporting. See the Anonymous usage statistics section.

Session Management:

• --session-idle-timeout-minutes: Session idle timeout in minutes. Sessions with no activity for this duration are automatically reaped - default: 30. Set to 0 to disable session reaping. Only relevant for SSE and streamable-http transports.

Tool Configuration:

• --enabled-tools: Comma-separated list of enabled categories - default: all categories except admin, agento11y, assistant, athena, clickhouse, cloudwatch, elasticsearch, examples, graphite, quickwit, runpanelquery, and snowflake. To enable disabled categories, add them to the list (e.g., "search,datasource,...,snowflake")
• --max-loki-log-limit: Maximum number of log lines returned per query_loki_logs call - default: 100. Note: Set this at least 1 below Loki's server-side max_entries_limit_per_query to allow truncation detection (the tool requests limit+1 internally to detect if more data exists).
• --loki-guardrail-mode: Loki query cost guardrail for query_loki_logs - default: off. Loki does not enforce max_query_bytes_read on log queries without a line filter, so a broad selector over a wide range can scan terabytes; the guardrail requires a selective stream selector, caps the effective time range (including range-vector durations like [30d]), and pre-checks Loki's index/stats byte estimate before running the query. shadow logs queries that would be blocked but lets them run (it still pays the index/stats round trip); enforce rejects them with rewrite guidance the LLM can act on. On VictoriaLogs the guardrail applies only to selector-shaped ({...}) queries — when no selector parses (the normal brace-less LogsQL shape), the query passes through entirely, and the byte-budget check never applies (no cheap index estimate). Env fallback: GRAFANA_LOKI_GUARDRAIL_MODE.
• --loki-guardrail-max-bytes: Maximum bytes a single query_loki_logs call may scan, estimated via Loki's index/stats API - default: 107374182400 (100 GiB). 0 disables the byte-budget check. Env fallback: GRAFANA_LOKI_GUARDRAIL_MAX_BYTES.
• --loki-guardrail-max-range: Maximum effective time range for a single query_loki_logs call, including range-vector durations - default: 24h. Accepts Go duration strings. 0 disables the range check. Env fallback: GRAFANA_LOKI_GUARDRAIL_MAX_RANGE.
• --loki-enforced-matchers: LogQL label matchers AND-ed into every native-Loki query to restrict which log streams can be read (e.g. environment=~"prod|staging"). Requires --disable-api. See Loki query enforcement.
• --loki-label-enumeration-fallback: What the label-enumeration tools do when negative enforced matchers can't scope them: reject (default) or unfiltered. See Loki query enforcement.
• --disable-search: Disable search tools
• --disable-datasource: Disable datasource tools
• --disable-incident: Disable incident tools
• --disable-prometheus: Disable prometheus tools
• --disable-write: Disable write tools (create/update operations)
• --disable-query: Disable query tools (tools that execute a query against a datasource); metadata and discovery tools stay available
• --enable-query: Keep the raw-SQL query tools (query_sql, query_influxdb) registered even under --disable-write. Equivalent to --enable-write-tools=query_sql,query_influxdb; kept as a shorthand for that common case.
• --enable-write-tools: Comma separated list of individual tool names to keep registered even under --disable-write, for tools whose write behavior is scoped enough to opt back in independently (e.g. find_error_pattern_logs,find_slow_requests). Has no effect on a tool whose whole category is disabled, e.g. via --disable-sift.
• --disable-loki: Disable loki tools
• --disable-elasticsearch: Disable elasticsearch and opensearch tools
• --disable-quickwit: Disable quickwit tools
• --disable-influxdb: Disable InfluxDB tools
• --disable-alerting: Disable alerting tools
• --disable-dashboard: Disable dashboard tools
• --disable-oncall: Disable oncall tools
• --disable-asserts: Disable asserts tools
• --disable-sift: Disable sift tools
• --disable-admin: Disable admin tools
• --disable-pyroscope: Disable pyroscope tools
• --disable-navigation: Disable navigation tools
• --disable-rendering: Disable rendering tools (panel/dashboard image export)
• --disable-snapshot: Disable snapshot tools
• --disable-cloudwatch: Disable CloudWatch tools
• --disable-examples: Disable query examples tools
• --disable-sql: Disable SQL datasource tools (ClickHouse, Snowflake, Athena, MySQL, PostgreSQL, MSSQL). Aliases --disable-clickhouse, --disable-snowflake, --disable-athena also work.
• --disable-runpanelquery: Disable run panel query tools
• --disable-graphite: Disable Graphite tools
• --disable-provisioning: Disable provisioning tools
• --disable-agento11y: Disable Agent Observability tools
• --disable-assistant: Disable Grafana Assistant tools
• --disable-docs: Disable documentation tools

Read-Only Mode

The --disable-write flag provides a way to run the MCP server in read-only mode, preventing any write operations to your Grafana instance. This is useful for scenarios where you want to provide safe, read-only access such as:

• Using service accounts with limited read-only permissions
• Providing AI assistants with observability data without modification capabilities
• Running in production environments where write access should be restricted
• Testing and development scenarios where you want to prevent accidental modifications

When --disable-write is enabled, the following write operations are disabled:

Dashboard Tools:

• update_dashboard

Folder Tools:

• create_folder

Incident Tools:

• create_incident
• add_activity_to_incident
• update_incident

Alerting Tools:

• alerting_manage_rules (create, update, delete operations)
• alerting_manage_silences (create, update, delete operations)

OnCall Tools:

• update_alert_group

Annotation Tools:

• create_annotation
• update_annotation
• delete_annotation

Sift Tools:

• find_error_pattern_logs (creates investigations)
• find_slow_requests (creates investigations)

These only create ephemeral Sift investigation records via the Sift API — they never touch a Grafana dashboard, alert, or datasource. Without them, list_sift_investigations/get_sift_investigation/get_sift_analysis have nothing to list or get. Pass --enable-write-tools=find_error_pattern_logs,find_slow_requests to keep them registered under --disable-write.

Snapshot Tools:

• create_snapshot
• delete_snapshot

Raw-SQL Query Tools:

These execute whatever query you give them without inspecting it, so they can write when the datasource credentials permit it — query_sql will run a DROP TABLE, query_influxdb will run a DELETE. Read-only mode therefore removes them. Pass --enable-query to keep them when the datasource credentials are known to be read-only.

• query_sql
• query_influxdb

Agent Observability Tools:

• agento11y_manage_evaluators (upsert, delete, fork, test evaluator operations)
• agento11y_manage_eval_rules (create, update, delete, preview rule and guard operations)
• agento11y_manage_eval_collections (save and delete saved conversations; create, update, delete collections; add and remove collection members)
• agento11y_manage_experiments (update and cancel experiment operations)
• agento11y_manage_test_suites (create and update test suites; create and publish versions; upsert and delete test cases)

All read operations remain available, allowing you to query dashboards, run PromQL/LogQL queries, list resources, and retrieve data. The query languages that cannot express a write — PromQL, LogQL, TraceQL, the Elasticsearch DSL, Graphite, CloudWatch — keep their query tools in read-only mode; only the raw-SQL ones listed above are removed.

Query-Free Mode

The --disable-query flag removes every tool that executes a query against a datasource, while leaving the metadata and discovery tools in place. This is useful when you want an assistant that can explore what exists — datasources, dashboards, metric names, labels, table schemas — without running potentially expensive or data-revealing queries, for example when the service account has datasources:read but not datasources:query.

It is the strongest of the three query settings, and it wins over --enable-query:

When --disable-query is enabled, the following tools are not registered:

Prometheus Tools:

• query_prometheus
• query_prometheus_histogram

Loki Tools:

• query_loki_logs
• query_loki_patterns

query_loki_stats and analyze_loki_labels stay registered: both send a selector to the datasource, but they read the index and return stream, chunk, and byte counts rather than log content.

Elasticsearch/OpenSearch and Quickwit Tools:

• query_elasticsearch
• query_quickwit

InfluxDB Tools (also removed by --disable-write, see above):

• query_influxdb

SQL Datasource Tools (also removed by --disable-write, see above):

• query_sql

Graphite Tools:

• query_graphite
• query_graphite_density

CloudWatch Tools:

• query_cloudwatch

Pyroscope Tools:

• query_pyroscope

Run Panel Query Tools:

• run_panel_query

The elasticsearch, quickwit, influxdb, and runpanelquery categories contain nothing else, so they register no tools at all when queries are disabled. The sibling tools in every other category — list_prometheus_metric_names, list_loki_label_values, describe_sql_table, list_cloudwatch_metrics, and so on — remain available.

Note that --disable-query gates the query tools and the grafana_api_request POST-to-/api/ds/query path, but does not police every route to a datasource. In read-only mode, grafana_api_request allows POST to /api/ds/query only when query tools are enabled (same gate as the raw-SQL tools — blocked by --disable-write unless --enable-query overrides). get_panel_image, which renders a panel server-side, is unaffected.

Client TLS Configuration (for Grafana connections):

• --tls-cert-file: Path to TLS certificate file for client authentication
• --tls-key-file: Path to TLS private key file for client authentication
• --tls-ca-file: Path to TLS CA certificate file for server verification
• --tls-skip-verify: Skip TLS certificate verification (insecure)

Server TLS Configuration (streamable-http transport only):

• --server.tls-cert-file: Path to TLS certificate file for server HTTPS
• --server.tls-key-file: Path to TLS private key file for server HTTPS

Usage

This MCP server works with both local Grafana instances and Grafana Cloud. For Grafana Cloud, use your instance URL (e.g., https://myinstance.grafana.net) instead of http://localhost:3000 in the configuration examples below.

1. If using service account token authentication, create a service account in Grafana with enough permissions to use the tools you want to use, generate a service account token, and copy it to the clipboard for use in the configuration file. Follow the [Grafana service account documentation][service-account] for details on creating service account tokens. Tip: If you're not comfortable configuring fine-grained RBAC scopes, a simpler (but less restrictive) option is to assign the built-in Editor role to the service account. This grants broad read/write access that covers most MCP server operations — use it when convenience outweighs strict least-privilege requirements.

   Note: The environment variable GRAFANA_API_KEY is deprecated and will be removed in a future version. Please migrate to using GRAFANA_SERVICE_ACCOUNT_TOKEN instead. The old variable name will continue to work for backward compatibility but will show deprecation warnings.

Reading the service account token from a file

Instead of passing the token inline via GRAFANA_SERVICE_ACCOUNT_TOKEN, you can point GRAFANA_SERVICE_ACCOUNT_TOKEN_FILE at a file path that contains the token. The file is read fresh on every request, so rotated tokens are picked up automatically without restarting the server.

This is particularly useful in Kubernetes, where a Secret mounted as a volume is updated in place when the underlying Secret changes (typically within ~1 minute). Combined with the per-request client cache — which is keyed on the token value — a rotated token transparently produces a new client with no pod restart and no downtime:

env:
  - name: GRAFANA_SERVICE_ACCOUNT_TOKEN_FILE
    value: /var/run/secrets/grafana/token
volumeMounts:
  - name: grafana-token
    mountPath: /var/run/secrets/grafana
    readOnly: true
volumes:
  - name: grafana-token
    secret:
      secretName: grafana-mcp-token

Surrounding whitespace (including a trailing newline) is trimmed from the file contents. If both GRAFANA_SERVICE_ACCOUNT_TOKEN and GRAFANA_SERVICE_ACCOUNT_TOKEN_FILE are set, the inline token takes precedence.

Multi-Organization Support

You can specify which organization to interact with using either:

• Environment variable: Set GRAFANA_ORG_ID to the numeric organization ID
• HTTP header: Set X-Grafana-Org-Id when using SSE or streamable HTTP transports (header takes precedence over environment variable - meaning you can set a default org as well).

When an organization ID is provided, the MCP server will set the X-Grafana-Org-Id header on all requests to Grafana, ensuring that operations are performed within the specified organization context.

Dynamic (per-call) organization selection

The options above fix the organization for the whole connection. To let a single connection target different organizations per tool call, start the server with the --dynamic-multi-org flag. This is off by default.

When enabled, every tool accepts an optional orgId argument that overrides the connection's organization for that call (driving both the X-Grafana-Org-Id header and, for app-platform APIs, the resolved Kubernetes namespace). Proxied datasource tools are additionally discovered across every organization the credential can access. Calls that omit orgId use the connection's default organization.

This only works for credentials that belong to more than one organization (e.g. a user or on-behalf-of identity); a service-account token remains bound to its single organization. Use the user_info tool to discover which orgId values are valid.

Example with organization ID:

{
  "mcpServers": {
    "grafana": {
      "command": "mcp-grafana",
      "args": [],
      "env": {
        "GRAFANA_URL": "http://localhost:3000",
        "GRAFANA_USERNAME": "<your username>",
        "GRAFANA_PASSWORD": "<your password>",
        "GRAFANA_ORG_ID": "2"
      }
    }
  }
}

Custom HTTP Headers

You can add arbitrary HTTP headers to all Grafana API requests using the GRAFANA_EXTRA_HEADERS environment variable. The value should be a JSON object mapping header names to values.

Example with custom headers:

{
  "mcpServers": {
    "grafana": {
      "command": "mcp-grafana",
      "args": [],
      "env": {
        "GRAFANA_URL": "http://localhost:3000",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<your token>",
        "GRAFANA_EXTRA_HEADERS": "{\"X-Custom-Header\": \"custom-value\", \"X-Tenant-ID\": \"tenant-123\"}"
      }
    }
  }
}

SOCKS5 Proxy

You can route all requests this server makes to Grafana through a SOCKS5 proxy using the GRAFANA_SOCKS5_PROXY environment variable. The proxy is scoped to this server's Grafana traffic: it does not modify the global HTTP_PROXY/HTTPS_PROXY variables, and when set it overrides their proxy selection for Grafana transports only, without affecting other MCP servers or your shell session. When unset, behavior is unchanged.

The URL must use the socks5:// or socks5h:// scheme (Go treats them identically: hostname resolution is delegated to the proxy) and may include credentials, e.g. socks5://user:pass@127.0.0.1:1080.

Example:

{
  "mcpServers": {
    "grafana": {
      "command": "mcp-grafana",
      "args": [],
      "env": {
        "GRAFANA_URL": "http://localhost:3000",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<your token>",
        "GRAFANA_SOCKS5_PROXY": "socks5://127.0.0.1:1080"
      }
    }
  }
}

An invalid proxy URL is a startup error, and if building a proxied connection fails at runtime the server fails closed rather than silently sending Grafana traffic directly.

Forwarding Headers from the Client (SSE/Streamable-HTTP Only)

When the MCP server runs behind a gateway or reverse proxy that handles SSO (e.g. an AWS ALB with OIDC), each user's session cookie must reach Grafana so it can associate the request with the authenticated user. The GRAFANA_FORWARD_HEADERS environment variable enables this by specifying a comma-separated allowlist of header names to copy from the incoming HTTP request to every outbound Grafana API request.

This only applies when using SSE (-t sse) or streamable-http (-t streamable-http) transports. It has no effect in stdio mode.

Example: forward the session cookie

{
  "env": {
    "GRAFANA_URL": "https://grafana.internal",
    "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<your token>",
    "GRAFANA_FORWARD_HEADERS": "Cookie"
  }
}

You can forward multiple headers by separating them with commas:

GRAFANA_FORWARD_HEADERS=Cookie,X-Session-Id

Forwarded headers are merged with any headers defined in GRAFANA_EXTRA_HEADERS. If a header name appears in both, the value from the incoming request takes precedence for that request.

Trace context headers (traceparent, tracestate, baggage) are the exception: the server propagates trace context itself, so a forwarded value never overrides the one it injects. See observability.

2. You have several options to install mcp-grafana:

   • uvx (recommended): If you have uv installed, no extra setup is needed — uvx will automatically download and run the server:

     bash

     uvx mcp-grafana
     

   • Docker image: Use the pre-built Docker image from Docker Hub.

     Important: The Docker image's entrypoint is configured to run the MCP server in SSE mode by default, but most users will want to use STDIO mode for direct integration with AI assistants like Claude Desktop:

     1. STDIO Mode: For stdio mode you must explicitly override the default with -t stdio and include the -i flag to keep stdin open:
     bash

     docker pull grafana/mcp-grafana
     # For local Grafana:
     docker run --rm -i -e GRAFANA_URL=http://localhost:3000 -e GRAFANA_SERVICE_ACCOUNT_TOKEN=<your service account token> grafana/mcp-grafana -t stdio
     # For Grafana Cloud:
     docker run --rm -i -e GRAFANA_URL=https://myinstance.grafana.net -e GRAFANA_SERVICE_ACCOUNT_TOKEN=<your service account token> grafana/mcp-grafana -t stdio
     

     Note — secure the networked modes: In SSE and streamable-http modes the container binds a non-loopback address (0.0.0.0:8000). Without a caller token the server starts but logs a security error (at the error log level, so it isn't hidden by --log-level; and it will refuse to start in a future major release). Set MCP_GRAFANA_SERVER_TOKEN to require an Authorization: Bearer <token> from clients (recommended). STDIO mode is unaffected. See Caller Authentication.

     2. SSE Mode: In this mode, the server runs as an HTTP server that clients connect to. You must expose port 8000 using the -p flag:
     bash

     docker pull grafana/mcp-grafana
     docker run --rm -p 8000:8000 -e GRAFANA_URL=http://localhost:3000 -e GRAFANA_SERVICE_ACCOUNT_TOKEN=<your service account token> -e MCP_GRAFANA_SERVER_TOKEN=<caller auth token> grafana/mcp-grafana
     

     3. Streamable HTTP Mode: In this mode, the server operates as an independent process that can handle multiple client connections. You must expose port 8000 using the -p flag: For this mode you must explicitly override the default with -t streamable-http
     bash

     docker pull grafana/mcp-grafana
     docker run --rm -p 8000:8000 -e GRAFANA_URL=http://localhost:3000 -e GRAFANA_SERVICE_ACCOUNT_TOKEN=<your service account token> -e MCP_GRAFANA_SERVER_TOKEN=<caller auth token> grafana/mcp-grafana -t streamable-http
     

     For HTTPS streamable HTTP mode with server TLS certificates:

     bash

     docker pull grafana/mcp-grafana
     docker run --rm -p 8443:8443 \
       -v /path/to/certs:/certs:ro \
       -e GRAFANA_URL=http://localhost:3000 \
       -e GRAFANA_SERVICE_ACCOUNT_TOKEN=<your service account token> \
       -e MCP_GRAFANA_SERVER_TOKEN=<caller auth token> \
       grafana/mcp-grafana \
       -t streamable-http \
       -addr :8443 \
       --server.tls-cert-file /certs/server.crt \
       --server.tls-key-file /certs/server.key
     

   • Download binary: Download the latest release of mcp-grafana from the releases page and place it in your $PATH.

   • Build from source: If you have a Go toolchain installed you can also build and install it from source, using the GOBIN environment variable to specify the directory where the binary should be installed. This should also be in your $PATH.

     bash

     GOBIN="$HOME/go/bin" go install github.com/grafana/mcp-grafana/cmd/mcp-grafana@latest
     

   • Deploy to Kubernetes using Helm: use the Helm chart from the Grafana helm-charts repository

     bash

     helm repo add grafana https://grafana.github.io/helm-charts
     helm install --set grafana.apiKey=<Grafana_ApiKey> --set grafana.url=<GrafanaUrl> my-release grafana/grafana-mcp
     

3. Add the server configuration to your client configuration file. For example, for Claude Desktop:

   If using uvx:

   json

   {
     "mcpServers": {
       "grafana": {
         "command": "uvx",
         "args": ["mcp-grafana"],
         "env": {
           "GRAFANA_URL": "http://localhost:3000",
           "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<your service account token>"
         }
       }
     }
   }
   

   If using the binary:

   json

   {
     "mcpServers": {
       "grafana": {
         "command": "mcp-grafana",
         "args": [],
         "env": {
           "GRAFANA_URL": "http://localhost:3000",  // Or "https://myinstance.grafana.net" for Grafana Cloud
           "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<your service account token>",
           // If using username/password authentication
           "GRAFANA_USERNAME": "<your username>",
           "GRAFANA_PASSWORD": "<your password>",
           // Optional: specify organization ID for multi-org support
           "GRAFANA_ORG_ID": "1"
         }
       }
     }
   }
   

Note: if you see Error: spawn mcp-grafana ENOENT in Claude Desktop, you need to specify the full path to mcp-grafana.

If using Docker:

{
  "mcpServers": {
    "grafana": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "-e",
        "GRAFANA_URL",
        "-e",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN",
        "grafana/mcp-grafana",
        "-t",
        "stdio"
      ],
      "env": {
        "GRAFANA_URL": "http://localhost:3000",  // Or "https://myinstance.grafana.net" for Grafana Cloud
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<your service account token>",
        // If using username/password authentication
        "GRAFANA_USERNAME": "<your username>",
        "GRAFANA_PASSWORD": "<your password>",
        // Optional: specify organization ID for multi-org support
        "GRAFANA_ORG_ID": "1"
      }
    }
  }
}

Note: The -t stdio argument is essential here because it overrides the default SSE mode in the Docker image.

Using VSCode with remote MCP server

If you're using VSCode and running the MCP server in SSE mode (which is the default when using the Docker image without overriding the transport), make sure your .vscode/settings.json includes the following:

"mcp": {
  "servers": {
    "grafana": {
      "type": "sse",
      "url": "http://localhost:8000/sse"
    }
  }
}

For HTTPS streamable HTTP mode with server TLS certificates:

"mcp": {
  "servers": {
    "grafana": {
      "type": "sse",
      "url": "https://localhost:8443/sse"
    }
  }
}

Debug Mode

You can enable debug mode for the Grafana transport by adding the -debug flag to the command. This will provide detailed logging of HTTP requests and responses between the MCP server and the Grafana API, which can be helpful for troubleshooting.

To use debug mode with the Claude Desktop configuration, update your config as follows:

If using the binary:

{
  "mcpServers": {
    "grafana": {
      "command": "mcp-grafana",
      "args": ["-debug"],
      "env": {
        "GRAFANA_URL": "http://localhost:3000",  // Or "https://myinstance.grafana.net" for Grafana Cloud
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<your service account token>"
      }
    }
  }
}

If using Docker:

{
  "mcpServers": {
    "grafana": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "-e",
        "GRAFANA_URL",
        "-e",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN",
        "grafana/mcp-grafana",
        "-t",
        "stdio",
        "-debug"
      ],
      "env": {
        "GRAFANA_URL": "http://localhost:3000",  // Or "https://myinstance.grafana.net" for Grafana Cloud
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<your service account token>"
      }
    }
  }
}

Note: As with the standard configuration, the -t stdio argument is required to override the default SSE mode in the Docker image.

TLS Configuration

If your Grafana instance is behind mTLS or requires custom TLS certificates, you can configure the MCP server to use custom certificates. The server supports the following TLS configuration options:

• --tls-cert-file: Path to TLS certificate file for client authentication
• --tls-key-file: Path to TLS private key file for client authentication
• --tls-ca-file: Path to TLS CA certificate file for server verification
• --tls-skip-verify: Skip TLS certificate verification (insecure, use only for testing)

Example with client certificate authentication:

{
  "mcpServers": {
    "grafana": {
      "command": "mcp-grafana",
      "args": [
        "--tls-cert-file",
        "/path/to/client.crt",
        "--tls-key-file",
        "/path/to/client.key",
        "--tls-ca-file",
        "/path/to/ca.crt"
      ],
      "env": {
        "GRAFANA_URL": "https://secure-grafana.example.com",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<your service account token>"
      }
    }
  }
}

Example with Docker:

{
  "mcpServers": {
    "grafana": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "-v",
        "/path/to/certs:/certs:ro",
        "-e",
        "GRAFANA_URL",
        "-e",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN",
        "grafana/mcp-grafana",
        "-t",
        "stdio",
        "--tls-cert-file",
        "/certs/client.crt",
        "--tls-key-file",
        "/certs/client.key",
        "--tls-ca-file",
        "/certs/ca.crt"
      ],
      "env": {
        "GRAFANA_URL": "https://secure-grafana.example.com",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<your service account token>"
      }
    }
  }
}

The TLS configuration is applied to all HTTP clients used by the MCP server, including:

• The main Grafana OpenAPI client
• Prometheus datasource clients
• Loki datasource clients
• Incident management clients
• Sift investigation clients
• Alerting clients
• Asserts clients

Direct CLI Usage Examples:

For testing with self-signed certificates:

./mcp-grafana --tls-skip-verify -debug

With client certificate authentication:

./mcp-grafana \
  --tls-cert-file /path/to/client.crt \
  --tls-key-file /path/to/client.key \
  --tls-ca-file /path/to/ca.crt \
  -debug

With custom CA certificate only:

./mcp-grafana --tls-ca-file /path/to/ca.crt

Programmatic Usage:

If you're using this library programmatically, you can also create TLS-enabled context functions:

// Using struct literals
tlsConfig := &mcpgrafana.TLSConfig{
    CertFile: "/path/to/client.crt",
    KeyFile:  "/path/to/client.key",
    CAFile:   "/path/to/ca.crt",
}
grafanaConfig := mcpgrafana.GrafanaConfig{
    Debug:     true,
    TLSConfig: tlsConfig,
}
contextFunc := mcpgrafana.ComposedStdioContextFunc(grafanaConfig)

// Or inline
grafanaConfig := mcpgrafana.GrafanaConfig{
    Debug: true,
    TLSConfig: &mcpgrafana.TLSConfig{
        CertFile: "/path/to/client.crt",
        KeyFile:  "/path/to/client.key",
        CAFile:   "/path/to/ca.crt",
    },
}
contextFunc := mcpgrafana.ComposedStdioContextFunc(grafanaConfig)

URL validation:

When calling NewGrafanaClient directly (stdio or programmatic construction), pre-validate URLs to avoid a reachable panic:

if err := mcpgrafana.ValidateGrafanaURL(urlFromHeader); err != nil {
    http.Error(w, err.Error(), http.StatusBadRequest)
    return
}
client := mcpgrafana.NewGrafanaClient(ctx, urlFromHeader, apiKey, nil)

Server TLS Configuration (Streamable HTTP Transport Only)

When using the streamable HTTP transport (-t streamable-http), you can configure the MCP server to serve HTTPS instead of HTTP. This is useful when you need to secure the connection between your MCP client and the server itself.

The server supports the following TLS configuration options for the streamable HTTP transport:

• --server.tls-cert-file: Path to TLS certificate file for server HTTPS (required for TLS)
• --server.tls-key-file: Path to TLS private key file for server HTTPS (required for TLS)

Note: These flags are completely separate from the client TLS flags documented above. The client TLS flags configure how the MCP server connects to Grafana, while these server TLS flags configure how clients connect to the MCP server when using streamable HTTP transport.

Example with HTTPS streamable HTTP server:

./mcp-grafana \
  -t streamable-http \
  --server.tls-cert-file /path/to/server.crt \
  --server.tls-key-file /path/to/server.key \
  -addr :8443

This would start the MCP server on HTTPS port 8443. Clients would then connect to https://localhost:8443/ instead of http://localhost:8000/.

Docker example with server TLS:

docker run --rm -p 8443:8443 \
  -v /path/to/certs:/certs:ro \
  -e GRAFANA_URL=http://localhost:3000 \
  -e GRAFANA_SERVICE_ACCOUNT_TOKEN=<your service account token> \
  grafana/mcp-grafana \
  -t streamable-http \
  -addr :8443 \
  --server.tls-cert-file /certs/server.crt \
  --server.tls-key-file /certs/server.key

Health Check Endpoint

When using the SSE (-t sse) or streamable HTTP (-t streamable-http) transports, the MCP server exposes a health check endpoint at /healthz. This endpoint can be used by load balancers, monitoring systems, or orchestration platforms to verify that the server is running and accepting connections.

Endpoint: GET /healthz

Response:

• Status Code: 200 OK
• Body: ok

Example usage:

# For streamable HTTP or SSE transport on default port
curl http://localhost:8000/healthz

# Probe a side listener while MCP stays on loopback (Kubernetes + sidecar)
./mcp-grafana -t streamable-http --address 127.0.0.1:8000 --healthz-address :8080
curl http://127.0.0.1:8080/healthz

Note: The health check endpoint is only available when using SSE or streamable HTTP transports. It is not available when using the stdio transport (-t stdio), as stdio does not expose an HTTP server.

Anonymous Usage Statistics

The server can report anonymous usage statistics about itself to Grafana Labs: which tools were called, how many of those calls failed, and how the server is configured. One report covers one server process — not one user and not one conversation — and is sent every 4h plus once on shutdown. Reporting is disabled by default in this release — the receiving endpoint isn't live yet — and a later release will change the default to enabled with the same opt-out.

Tool arguments, resource names, queries, log lines, error messages and credentials are never sent. Flags are recorded by name only, never by value, and the Grafana instance is described only as cloud or self_hosted — never by URL, hostname, stack slug or org. Nothing is per user, per session or per client: there is no session identifier on the wire and no way to attribute a tool call to a particular client.

# Turn reporting on
mcp-grafana --usage-stats=enabled

# Turn it off (or GRAFANA_USAGE_STATS=disabled)
mcp-grafana --usage-stats=disabled

# Print what would be sent, to stderr, and send nothing
GRAFANA_USAGE_STATS=log mcp-grafana

DO_NOT_TRACK=1 also disables reporting, following the cross-tool DO_NOT_TRACK convention. Only 1 has any effect, it can only disable, and both --usage-stats and GRAFANA_USAGE_STATS override it, so a host that sets it globally can still opt one server back in.

GRAFANA_USAGE_STATS_ENDPOINT changes the destination. It is not an opt-out.

For the full field list, what is never sent, how to read the data and its limitations, see Anonymous usage statistics.

Observability

The MCP server supports Prometheus metrics, OpenTelemetry distributed tracing, and OpenTelemetry log export, following the OTel MCP semantic conventions. Tracing and log export are configured via standard OTEL_* environment variables and work with any transport.

Note: mcp-grafana currently only supports the OTLP/gRPC transport for both traces and logs. OTEL_EXPORTER_OTLP_PROTOCOL (and its _TRACES_PROTOCOL / _LOGS_PROTOCOL variants) are not honored — gRPC is used regardless.

Metrics

When using the SSE or streamable HTTP transports, enable Prometheus metrics with the --metrics flag:

# Metrics served on the main server at /metrics
./mcp-grafana -t streamable-http --metrics

# Metrics served on a separate address
./mcp-grafana -t streamable-http --metrics --metrics-address :9090

Available Metrics:

Note: Metrics are only available when using SSE or streamable HTTP transports. They are not available with the stdio transport.

When the Loki cost guardrail (--loki-guardrail-mode) is enabled, four more counters record its decisions:

reason is one of selector, range, bytes; cause is one of unparseable, estimate_failed; backend is one of loki, victorialogs, unknown. A query that trips several checks is counted once, labelled with the check that ran first (selector, then range, then bytes), so the four counters partition the guarded population. See Observability for how to read them during a shadow → enforce rollout.

Library embedders should set GrafanaConfig.MeterProvider (the metrics counterpart of GrafanaConfig.Logger): the guardrail runs inside a tool handler, so it has no constructor option, and a process that installs a noop global MeterProvider would otherwise drop every recording.

Slow-request logging

The --slow-request-threshold flag emits a structured log event whenever an MCP request (tool invocation, list, resource read, etc.) exceeds the given duration. It is useful for diagnosing slow queries and tool calls without drowning in the full debug log.

# Warn on any request slower than 500ms (works on all transports)
./mcp-grafana -t streamable-http --slow-request-threshold 500ms

# Same thing on stdio (the feature is transport-agnostic, unlike --metrics)
./mcp-grafana -t stdio --slow-request-threshold 500ms

# Log at INFO level instead of WARN (useful during investigation)
./mcp-grafana -t streamable-http --slow-request-threshold 500ms --slow-request-log-level info

The log event carries these structured attributes:

Slow-request logging works on all transports (including stdio) and does not require --metrics. The default threshold of 0 disables it entirely. Proxied tools flow through tools/call and are covered automatically.

Tracing

Distributed tracing is configured via standard OTEL_* environment variables and works independently of the --metrics flag. When OTEL_EXPORTER_OTLP_ENDPOINT (or the signal-specific OTEL_EXPORTER_OTLP_TRACES_ENDPOINT) is set, the server exports traces via OTLP/gRPC:

# Send traces to a local Tempo instance
OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317 \
OTEL_EXPORTER_OTLP_INSECURE=true \
./mcp-grafana -t streamable-http

# Send traces to Grafana Cloud with authentication
OTEL_EXPORTER_OTLP_ENDPOINT=https://tempo-us-central1.grafana.net:443 \
OTEL_EXPORTER_OTLP_HEADERS="Authorization=Basic ..." \
./mcp-grafana -t streamable-http

Tool call spans follow semconv naming (tools/call <tool_name>) and include attributes like gen_ai.tool.name, mcp.method.name, and mcp.session.id. The server also supports W3C trace context propagation from the _meta field of tool call requests.

Logs

When OTEL_EXPORTER_OTLP_ENDPOINT (or the signal-specific OTEL_EXPORTER_OTLP_LOGS_ENDPOINT) is set, the server also exports structured logs via OTLP/gRPC in addition to the existing plain-text stderr output. The otelslog bridge automatically attaches trace_id and span_id from the active span, so log records correlate with the traces the server already emits.

Traces and logs resolve their endpoints independently, so the two signals can be enabled separately: setting only OTEL_EXPORTER_OTLP_TRACES_ENDPOINT enables tracing without log export, setting only OTEL_EXPORTER_OTLP_LOGS_ENDPOINT enables log export without tracing, and the generic OTEL_EXPORTER_OTLP_ENDPOINT enables both.

If you use the generic OTEL_EXPORTER_OTLP_ENDPOINT but want to disable log export (e.g. your backend does not support the LogsService), set:

OTEL_LOGS_EXPORTER=none

This prevents the server from creating an OTLP logs exporter regardless of the endpoint configuration, avoiding errors like unknown service opentelemetry.proto.collector.logs.v1.LogsService.

Stderr logging is unchanged when OTLP logging is enabled; you can continue to rely on container logs or pipe stderr to /dev/null if you prefer.

# Send both logs and traces to a local OTel collector
OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317 \
OTEL_EXPORTER_OTLP_INSECURE=true \
./mcp-grafana -t streamable-http

The transport is OTLP/gRPC (default port 4317). Logs can be sent directly to any managed backend that accepts OTLP/gRPC — for example, Grafana Cloud — by pointing OTEL_EXPORTER_OTLP_LOGS_ENDPOINT (or the generic OTEL_EXPORTER_OTLP_ENDPOINT) at the remote gRPC endpoint and supplying auth via OTEL_EXPORTER_OTLP_LOGS_HEADERS (or OTEL_EXPORTER_OTLP_HEADERS), mirroring the tracing example above. A local OTel collector is optional — useful for fan-out, batching, or multi-backend routing, but not required.

The signal-specific variants OTEL_EXPORTER_OTLP_LOGS_ENDPOINT, OTEL_EXPORTER_OTLP_LOGS_HEADERS, OTEL_EXPORTER_OTLP_LOGS_INSECURE, OTEL_EXPORTER_OTLP_LOGS_CERTIFICATE, OTEL_EXPORTER_OTLP_LOGS_TIMEOUT, and OTEL_EXPORTER_OTLP_LOGS_COMPRESSION are honored and override their generic OTEL_EXPORTER_OTLP_* counterparts — see the OTel exporter spec for the full list and precedence rules.

If the configured collector is unreachable, log records are buffered in memory (default queue: 2048) and the oldest records are dropped once the queue fills. The process continues without blocking the service. Configure a local OTel collector if you need lossless buffering during outages.

Logs are also exported under the stdio transport, which makes it easy to centralize logs from local mcp-grafana instances invoked by IDE clients.

Docker example with metrics, tracing, and logs:

docker run --rm -p 8000:8000 \
  -e GRAFANA_URL=http://localhost:3000 \
  -e GRAFANA_SERVICE_ACCOUNT_TOKEN=<your token> \
  -e OTEL_EXPORTER_OTLP_ENDPOINT=http://tempo:4317 \
  -e OTEL_EXPORTER_OTLP_INSECURE=true \
  grafana/mcp-grafana \
  -t streamable-http --metrics

Loki query enforcement

--loki-enforced-matchers lets an operator restrict which Loki log streams the server can ever read, by AND-ing a fixed set of LogQL label matchers into every native-Loki query the server issues. This is useful when a datasource contains streams that must not be exposed (e.g. logs that may carry se