For the complete documentation index, see llms.txt. Markdown versions are available by appending .md to documentation URLs.

SigNoz MCP Server

SigNoz Cloud - This page applies to SigNoz Cloud editions.
Self-Host - This page applies to self-hosted SigNoz editions.

The SigNoz MCP Server implements the Model Context Protocol (MCP), an open standard that lets AI agents interact with your SigNoz observability data. Use natural language to query metrics, traces, and logs; investigate alerts; and manage dashboards, alert rules, saved views, and notification channels.

The MCP server works best with the SigNoz plugin: one install adds the server registration and every Agent Skill, so your agent can query, build, and troubleshoot in SigNoz from the first prompt.

Connect to SigNoz's MCP server

Connect your AI tool to SigNoz Cloud's hosted MCP server. No installation required: just add the URL and authenticate.

https://mcp.<region>.signoz.cloud/mcp

Verify these values:

  • <region>: Your SigNoz Cloud region. Find your region under Settings → Ingestion in SigNoz, or see the region reference.

Install in one click

Manual Configuration

Add this configuration to .cursor/mcp.json:

.cursor/mcp.json
{
  "mcpServers": {
    "signoz": {
      "url": "https://mcp.<region>.signoz.cloud/mcp"
    }
  }
}

Verify these values:

  • <region>: Your SigNoz Cloud region

Need help? See the Cursor MCP docs ↗

Authenticate from your client

When you add the hosted MCP URL to your client, the MCP client will initiate an authentication flow. You will be prompted to enter:

  1. Your SigNoz instance URL (e.g., https://your-instance.signoz.cloud)
  2. Your API key: go to Settings → Service Accounts in SigNoz, create a service account, and generate an API key (requires Admin role). See Service Accounts for details.

Header-Based Authentication

Some MCP clients (like Cursor Automations) run entirely in the cloud and cannot perform interactive OAuth authentication. For these clients, use header-based authentication by passing your API key and instance URL directly in the request headers.

mcp.json
{
  "mcpServers": {
    "signoz": {
      "url": "https://mcp.<region>.signoz.cloud/mcp",
      "headers": {
        "SIGNOZ-API-KEY": "<your-api-key>",
        "X-SigNoz-URL": "<your-signoz-instance-url>"
      }
    }
  }
}

Verify these values:

  • <region>: Your SigNoz Cloud region
  • <your-api-key>: API key from SigNoz Service accounts
  • <your-signoz-instance-url>: Your SigNoz Cloud instance URL

Validate

After connecting, verify the server is working:

  1. Open your AI agent.
  2. Ask: "List all alerts" or "Show me all available services".
  3. The agent should return structured results from your SigNoz instance.
  4. Ask for a dashboard or trace link and open the returned link. For Self-Hosted SigNoz, configure the browser URL if the link contains an internal hostname.

Available Tools

The MCP server exposes the following tools to your AI agent:

ToolDescription
signoz_get_org_overviewGet the current status and overall posture of your SigNoz deployment, including the org's feature flags
signoz_list_metricsDiscover active metric names and catalog metadata such as type, temporality, and unit
signoz_query_metricsQuery a known metric for values, trends, breakdowns, or formulas with metric-aware defaults
signoz_get_top_metricsReturn top 100 metrics ranked by ingested sample volume with percentages for cost and volume analysis
signoz_check_metric_usageGiven a list of metric names (up to 50 per call), return which dashboards and alerts reference each one
signoz_check_metric_cardinalityReturn label/attribute keys for a single metric with cardinality counts and sample values, sorted highest-cardinality first
signoz_get_field_keysDiscover available field keys for metrics, traces, or logs, including AI trace fields
signoz_get_field_valuesGet possible values for a field key
signoz_list_alertsList firing, silenced, or inhibited Alertmanager alert instances
signoz_list_alert_rulesList configured alert rules, including inactive or disabled rules; filter by name, label, severity, or state, and sort before paging
signoz_get_alertGet an alert rule definition by ID
signoz_get_alert_historyGet a rule's firing and state-transition history (cursor-paginated)
signoz_create_alertCreate a threshold, PromQL, or metric anomaly alert rule in the v2alpha1 schema, with direct or policy routing
signoz_update_alertUpdate an existing alert rule
signoz_delete_alertDelete an alert rule
signoz_list_dashboardsList all dashboards with summaries
signoz_get_dashboardGet full dashboard configuration
signoz_create_dashboardCreate a new dashboard
signoz_update_dashboardReplace a dashboard with the complete edited definition from signoz_get_dashboard
signoz_patch_dashboardApply a partial JSON Patch to a dashboard without resending the whole definition
signoz_delete_dashboardDelete a dashboard by UUID
signoz_list_dashboard_templatesSearch the bundled SigNoz dashboard template catalog by technology, such as postgres or kubernetes pod, so the model can pick one
signoz_import_dashboardCreate a dashboard from a curated SigNoz dashboard template by path
signoz_list_servicesList APM services with trace activity in a time range
signoz_get_service_top_operationsGet top operations for a service
signoz_list_viewsList saved views for Logs, Traces, Metrics, or Cost Meter
signoz_get_viewGet a saved Explorer view by UUID
signoz_create_viewCreate a saved Explorer view
signoz_update_viewReplace an existing saved Explorer view
signoz_delete_viewDelete a saved Explorer view
signoz_aggregate_logsAggregate logs with Query Builder expressions such as count(), p99(duration), or countIf(...), with grouping and having
signoz_search_logsSearch logs with flexible filtering; selectFields adds fields to each row
signoz_aggregate_tracesAggregate span statistics, or per-trace cost, token, and call counts for AI traces, with grouping and having
signoz_search_tracesSearch span rows, return one row per trace (including AI traces with cost and token counts), or discover trace IDs; selectFields adds fields to span rows
signoz_get_trace_detailsGet a full trace with all spans by trace ID alone; no time range needed
signoz_get_ai_pricingList the AI model prices SigNoz uses to compute span cost, or models seen in the last hour that have no price
signoz_list_ai_attribute_mappingsList the attribute mapping groups that turn vendor AI span attributes into gen_ai.* keys
signoz_test_ai_attribute_mappingRun sample spans through the AI attribute mappings without saving anything
signoz_execute_builder_queryRun Query Builder v5 requests the dedicated tools cannot express, including multi-query requests, formulas, PromQL, and ClickHouse SQL
signoz_search_docsSearch official SigNoz docs for product, setup, instrumentation, config, API, deployment, or troubleshooting questions
signoz_fetch_docFetch full markdown for one official SigNoz docs page or heading
signoz_list_notification_channelsList channel summaries without provider configuration
signoz_get_notification_channelGet one channel's full configuration by ID
signoz_create_notification_channelCreate a notification channel; optionally send a test notification
signoz_update_notification_channelReplace a notification channel's full configuration; optionally send a test notification
signoz_delete_notification_channelDelete a notification channel by ID

The server also exposes read-only MCP resources (signoz://... URIs) with schemas, query-builder guides, and payload examples that AI agents read before composing queries and alert, dashboard, view, and notification-channel payloads.

For LLM and agent questions, such as cost per trace or tool-call counts, signoz_search_traces, signoz_aggregate_traces, and signoz_get_field_keys accept queryType: builder_ai_query, the query type the AI Traces Query API uses.

Agents should discover fields before filtering logs or traces, and reuse the current resource definition when making a replacement update. Notification-channel writes default to test=false; use test=true only when a test notification is requested. Tools that return webUrl provide a ready-to-open SigNoz link, so agents can return it directly.

For detailed parameter reference and the full resource list, see the GitHub repository README.

Troubleshooting

Tools not appearing in the AI client

  • Verify the binary path is correct and the file is executable (self-hosted only).
  • Set LOG_LEVEL=debug for detailed server logs.

Authentication errors

  • For SigNoz Cloud: ensure you are using the correct region in the MCP URL.
  • Confirm your API key is valid and has not expired.
  • For self-hosted HTTP mode where the client sends the key, use the SIGNOZ-API-KEY: <your-api-key> header.
  • UNAUTHORIZED means the credentials need attention. For PERMISSION_DENIED, check the service account's role and permissions for the requested operation.

SigNoz Cloud connection issues

  • Verify the region in the URL matches your account (https://mcp.<region>.signoz.cloud/mcp).
  • Check your network can reach mcp.<region>.signoz.cloud.
  • Make sure you completed the authentication flow when prompted.

Server not starting (self-hosted HTTP mode)

  • Ensure port 8000 is not already in use.
  • Check that TRANSPORT_MODE=http is set.
  • Review server logs for startup errors.

Configure global.external_url on the SigNoz server as described in Configure browser links, then restart SigNoz. The MCP server may keep a previously discovered URL for up to 15 minutes. Check MCP logs for global-config discovery warnings if links still fall back to SIGNOZ_URL.

Is this page helpful

Last updated—November 03, 2026