Skip to main content
Glama
sifflet-mcp

sifflet-mcp

Official
by siffletdata

Sifflet MCP Server

An MCP (Model Context Protocol) server that enables data observability operations with the Sifflet platform.

Features

This project provides an MCP server enabling interactions with Sifflet API :

  • Explore assets: Search for tables, views, dashboards, and other data assets. View their schema, owners, tags, and their metadata.

  • Explore monitors: Discover existing monitors and generate their Monitor-as-Code YAML configurations.

  • Generate new monitors from a description: turn a plain-English requirement (e.g. "alert when row count drops below 1000 on the 'orders' table") into a Monitor-as-Code YAML snippet for a given list of datasets. Requires Editor role on the targeted domain.

  • Explore incidents: List all data observability incidents detected by the Sifflet platform.

  • Perform impact analysis: Start from an incident and trace the downstream assets affected.

Related MCP server: Datadog MCP Server

Sample Use Cases

Here are a few scenarios where the Sifflet MCP Server can be particularly helpful:

  • Understanding Downstream Impact: You're modifying a dbt model and need to identify the owners of dependent downstream models and dashboards. The MCP server can provide these details, allowing you to proactively notify them about your upcoming changes.

  • Accessing Up-to-Date Table Metadata: You're about to update a table in your data warehouse. Before you proceed, you can query the MCP server to get its latest metadata. This includes information on how the table is currently monitored in Sifflet, whether it's involved in any ongoing incidents, the list of its frequent users, and other relevant operational details.

  • Bootstrapping New Asset Monitoring: You're creating a new table (or dbt model) and want to ensure it's well-monitored from the start. You can ask the MCP server to list the Sifflet monitors already created for similar existing assets. The server can then provide the Monitor-as-Code YAML configurations, which you can adapt and deploy.

  • Generating a Monitor from a Description: You want to export the YAML configuration of a Monitor you want to create (e.g. "alert when row count drops below 1000 on the 'orders' table"). The get_monitor_code_by_description tool returns a Monitor-as-Code YAML snippet for a given list of datasets that you can adapt and commit. Note: requires Editor role on the targeted domain (see Prerequisites).

Usage

Prerequisites

  • uv (Python package installer/environment manager)

      # uv installation script for Linux/MacOS
      curl -LsSf https://astral.sh/uv/install.sh | sh
  • A Sifflet backend running locally or remotely. You will need the following information:

    • SIFFLET_API_TOKEN: see how to generate one. A token with the Viewer role is enough for most tools. The get_monitor_code_by_description tool additionally requires Editor role on the targeted domain. If you plan to use that tool, generate a token with Editor access on that domain.

    • SIFFLET_BACKEND_URL: Full URL to the Sifflet backend for instance: https://<tenant_name>.siffletdata.com/api/

Using with MCP Clients

Cursor

Add the following configuration in the mcp.json. Follow Cursor instructions to set it up.

{
  "mcpServers": {
    "mcp_server_sifflet": {
      "command": "uvx",
      "args": ["sifflet-mcp@latest"],
      "env": {
        "SIFFLET_API_TOKEN": "<access_token>",
        "SIFFLET_BACKEND_URL": "https://<tenant_name>.siffletdata.com/api/"
      }
    }
  }
}

Note: You may need to use the full path to the uvx executable in the command field. You can find the full path by running which uvx in your terminal.

Claude Desktop

Follow the instructions in the Claude documentation to set up claude_desktop_config.json.

Then, add the following configuration to your claude_desktop_config.json file:

{
  "mcpServers": {
    "sifflet-mcp": {
      "command": "uvx",
      "args": ["sifflet-mcp@latest"],
      "env": {
        "SIFFLET_API_TOKEN": "<access_token>",
        "SIFFLET_BACKEND_URL": "https://<tenant_name>.siffletdata.com/api/"
      }
    }
  }
}

Note: You may need to use the full path to the uvx executable in the command field. You can find the full path by running which uvx in your terminal.

Contributing

For development setup and contribution guidelines, please see CONTRIBUTING.md.

Reporting Problems

If you encounter any problems or have a bug to report, please feel free to open an issue on this GitHub repository. Alternatively, you can reach out to your Sifflet Customer Success team.

Available Tools

10 tools
asset_by_urnA
    Get asset information by urn. The urn is the unique identifier a asset, for example dataset:0826ce5c-7027-4857-aa47-b639265d1867. It can be found when you search for an asset.
    

Inputs: asset_urn (required).

close_incident_by_idA
    Close an incident by id. The id is the unique identifier of the incident, for example 1234.
    The should_qualify_monitor is a boolean that indicates if the monitor should be qualified or not.
    

Inputs: incident_id (required), should_qualify_monitor (required).

get_downstream_assets_of_assetA
      Get all downstream assets of an asset. An Urn is the unique identifier of an asset, for example dataset:0826ce5c-7027-4857-aa47-b639265d1867. It can be found when you search for an asset.
      

Inputs: urn (required).

get_incident_by_issue_numberA

Get incident details by issue number. The issue number is the unique identifier of the incident, for example 1234.

Inputs: issue_nbr (required).

get_incident_scope_by_issue_numberA

Get incident scope by issue number. The issue number is the unique identifier of the incident, for example 1234. The scope is the list of assets and monitors associated with the incident.

Inputs: issue_number (required).

get_monitor_code_by_descriptionA
    Returns a monitor configuration based on the input query
    The input is a description of the monitor you want to create.
    The description should be a natural language description of the monitor you want to create.
    The dataset_ids is a list of dataset ids that the monitor should be applied to. You must provide at least one dataset id. You can find the id of the dataset using search_asset. 
    The description is a string that describes the monitor you want to create. For example: "Create a monitor that checks if the number of rows in the table is greater than 1000".
    The result is a YAML string that contains the monitor configuration.
    

Inputs: dataset_ids (required), description (required).

get_monitor_details_by_idA
    Get monitor details by id. 
    The monitor_id is the monitor UUID. The id can be found when you search for a monitor.
    monitor_id is the unique identifier of the monitor, for example 3b8a1333-aef0-468a-893b-e87dfc095c82
    

Inputs: monitor_id (required).

open_incident_by_idB

Open an incident by id. The id is the unique identifier of the incident, for example 1234.

Inputs: incident_id (required).

search_assetA
    Search assets, tables, dashboards, pipelines.
    asset_type can be one of the following: TABLE_AND_VIEW, PIPELINE, DASHBOARD, ML_MODEL
    health_status can be one of the following: URGENT_INCIDENTS, HIGH_RISK_INCIDENTS, NO_INCIDENTS, NOT_MONITORED, UNSUPPORTED
    owners is a list of owners emails associated with the asset
    tags is a list of tags associated with the asset
    text_search is a string to search in the asset name or description
    items_per_page is the number of items to return per page
    page is the page number to return. Pages start at 0.
    

Inputs: page (required), tags (required), asset_type (required), text_search (required), +3 more.

search_incidentsA
      Search incidents
      Sort is done by createdAt by default. By default sort is descending. But you can change it using sort=asc.
      You can filter by status, user and text search.
      Status can be OPEN, IN_PROGRESS, CLOSED.
      Text search is a string.
      Pages start at 0.
      

Inputs: page (required), status (required), text_search (required), items_per_page (required), +1 more.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.6/5.0

    Scored across 10 tools

    Disambiguation4/5

    Most tools have clearly distinct purposes (asset lookup vs incident search vs monitor details). However, incident tools use different identifiers ('by_id' vs 'by_issue_number') for what appears to be the same entity, which could confuse an agent. Asset tools by URN are distinct but share similar input patterns.

    Naming Consistency3/5

    Naming conventions are mixed: 'asset_by_urn' does not follow the verb-first pattern used by most others, and pluralization is inconsistent ('search_asset' vs 'search_incidents'). Incident tools also inconsistently use 'by_id' versus 'by_issue_number'.

    Tool Count5/5

    10 tools is well-scoped for a data observability platform, covering assets, monitors, and incidents without being overwhelming. Each tool serves a distinct function, and the count fits comfortably within the ideal range.

    Completeness3/5

    Core workflows like asset search/detail, incident management, and monitor code generation are covered. However, there is no tool to search or list monitors (only get by ID), no monitor creation/update, and only downstream asset lineage (no upstream). These are notable gaps for a monitoring-focused tool.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      Not graded
      quality
      D
      maintenance
      Enables interaction with the Bigeye Data Observability platform to query data quality issues, analyze data lineage and dependencies, track AI agent data access patterns, and perform root cause analysis for data issues.
      -
    • F
      license
      B
      quality
      F
      maintenance
      Enables comprehensive Datadog monitoring capabilities including CI/CD pipeline management, service logs analysis, metrics querying, monitor and SLO management, service definitions retrieval, and team management through Claude and other MCP clients.
      13
      20
      -
    • A
      license
      B
      quality
      C
      maintenance
      Enables LLM agents to query Dynatrace SaaS for observability data (logs, metrics, traces, entities, problems, vulnerabilities) and manage configurations (dashboards, notebooks, SLOs, synthetic monitors, settings).
      100
      MIT
    • A
      license
      Not graded
      quality
      D
      maintenance
      Enables AI assistants to interact with Datadog's observability platform via natural language, covering metrics, logs, APM, monitors, dashboards, incidents, and infrastructure.
      1
      MIT