Monoscope MCP Server

What you get

• REST tools — every operation in the public API surfaces as a verb-first MCP tool: list_monitors, get_monitor, mute_monitor, search_events, apply_dashboard, whoami, get_schema, … About 50 in total. New endpoints become MCP tools automatically; renames don’t require a server release.
• Workflow tools — bundled actions that combine several internal calls into one agent-friendly verb: - find_error_patterns — top established log patterns ranked by current-hour count. Use to answer “what is blowing up right now” without crafting a query. - search_events_nl — translate a natural-language description into a KQL query (returns the query for the agent to execute via search_events). - analyze_issue — fetch an issue and return both the payload and an LLM-generated diagnosis (probable cause / signals / next steps).
• Same auth as the REST API — Bearer API key. Project scope is encoded in the key; one key per project. See API Reference → Getting Started for how to mint one.

Quick start

{
  "mcpServers": {
    "monoscope": {
      "url": "https://api.monoscope.tech/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

{
  "mcpServers": {
    "monoscope": {
      "url": "https://api.monoscope.tech/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Verifying with curl

# List tools
curl -s -H "Authorization: Bearer $MONOSCOPE_API_KEY" \
     -H 'Content-Type: application/json' \
     -X POST https://api.monoscope.tech/api/v1/mcp \
     -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' \
     | jq '.result.tools[].name' | head -10

# Call a tool
curl -s -H "Authorization: Bearer $MONOSCOPE_API_KEY" \
     -H 'Content-Type: application/json' \
     -X POST https://api.monoscope.tech/api/v1/mcp \
     -d '{
           "jsonrpc": "2.0",
           "id": 2,
           "method": "tools/call",
           "params": {"name": "find_error_patterns", "arguments": {"limit": 10}}
         }' \
     | jq '.result.structuredContent.patterns[0]'

Tool catalog

Naming convention

Verb-first snake_case throughout: list_X, get_X, create_X, <action>_<resource> (e.g. mute_monitor). No {id} placeholders in tool names — those live in the inputSchema. This matches the conventions used by Sentry, Grafana, Datadog, and Honeycomb’s MCP servers.

Tool result envelope

Every tool call returns the standard MCP envelope:

{
  "content": [{"type": "text", "text": "<JSON or message>"}],
  "isError": false,
  "structuredContent": { ... }
}

• content — text fallback (the JSON body or an error message). Always present.
• isError — true when the underlying handler returned a non-2xx status (or the workflow tool reported a problem). Never raised as a JSON-RPC error so the agent can react to it.
• structuredContent — the parsed JSON result when the response was JSON. Use this for typed handling.

Auth, project scoping, and security

• Bearer token in the Authorization header, same scheme as the REST API. Mint keys at Settings → API Keys in the dashboard, or via monoscope api-keys create from the CLI.
• A key is scoped to one project. The MCP server reads the project from the key — no X-Project-Id header to remember.
• The endpoint is hosted on TLS-only api.monoscope.tech. Do not put a raw API key into a client config that other people share — use per-user keys.
• The MCP route does not expose itself as a tool, so an agent cannot recursively call /api/v1/mcp from inside a tool call.

See also

• API Reference — the same operations as raw HTTP.
• CLI & Agents — terminal access to the same surface.
• Claude Code Skills — natural-language skills that drive the CLI; complementary to the MCP server.