# Coalesce MCP for AI Agents MCP > Coalesce MCP gives your AI agent direct control over Snowflake data pipelines. It lets you list all environments, check job status, and trigger complex transformations right from chat. You manage your entire ETL workflow without touching a UI. ## Overview - **Category:** developer-tools - **Price:** Free - **Tags:** data-transformation, snowflake, etl-pipeline, data-engineering, job-monitoring, column-aware ## Description This MCP connects your AI client directly to Coalesce, the platform that manages data transformation for Snowflake. Instead of opening multiple dashboards or writing boilerplate API calls, you talk to your agent and tell it exactly what needs transforming. Need to know if yesterday's run failed in staging? Ask your agent; it checks the job status immediately. Want to test a new pipeline on demand? Your agent triggers that specific transformation for you. You can list out every environment configured, from development through production, and inspect their current settings. It’s all about making data governance visible via natural language. If you're working with complex Snowflake pipelines, this MCP lets your AI client manage those transformations and monitor jobs without needing the Coalesce UI open. It integrates into your existing toolset; just connect it through Vinkius and let your agent handle the heavy lifting. ## Tools ### get_environment Retrieves detailed configuration information for a specific data pipeline environment. ### get_job_details Fetches comprehensive details about a particular job, including its historical performance and settings. ### get_run_status Checks the current progress or final status of any triggered data pipeline run. ### list_environments Retrieves a comprehensive list of all environments configured in your Coalesce organization. ### list_jobs Gets a roster of available jobs, with the option to filter them by which environment they belong to. ### list_nodes Retrieves metadata about specific transformation nodes within an active environment. ### trigger_job Manually starts a predefined data transformation job inside a specified environment. ### trigger_run Initiates a brand new, full run for an entire environment, optionally targeting a specific job within it. ## Prompt Examples **Prompt:** ``` I need a status update on my Production data warehouse pipelines. ``` **Response:** ``` **Production Pipeline Status Report** * **Environment:** Production (env-001) * **Last Run Status:** ✅ Success, 2 hours ago * **Current Activity:** All core dimension tables refreshed successfully. * **Next Steps:** Fact nodes are currently materializing in Snowflake. Estimated completion: 30 minutes. ``` **Prompt:** ``` List all my data environments and tell me which one failed recently. ``` **Response:** ``` You have three configured environments: 1. 'Production' (env-001) — Last run: success, 2h ago 2. 'Staging' (env-002) — Last run: success, 5h ago 3. 'Development' (env-003) — **Last run: failed**, 1 hour ago. Do you want me to retrieve the detailed error logs for Development? ``` **Prompt:** ``` Start a test run in Staging using only the sales node. ``` **Response:** ``` Okay, initiating job 'sales-etl' in the Staging environment. Run ID: `run-8k3m`. Status: Running. * **Target:** Sales Node Metadata retrieved successfully. * **Progress:** 10% complete (Initializing connection). * **Expected Time:** Approximately 4 minutes. ``` ## Capabilities ### List all configured environments It retrieves a complete list of every development, staging, or production environment set up in your Coalesce organization. ### Get specific environment details You can pull detailed configurations for any single environment to verify settings before making changes. ### Check job run status and logs Your agent checks the current progress of a pipeline run, providing real-time updates or viewing failure logs. ### List all available jobs It provides an inventory of all data transformation jobs, allowing you to filter by environment or job type. ### Trigger a new pipeline run You tell your agent which environment and what job to use, and it starts the required data transformation immediately. You can also specify nodes to narrow the scope of the run. ## Use Cases ### Needing to check if Production data is ready for launch A team lead needs confirmation that the latest dataset finished running correctly. They ask their agent: 'What's the status of the production pipeline?' The agent runs `get_run_status` and confirms, 'The run is at 98%, all fact nodes materialized.' This prevents last-minute deployment errors. ### Debugging a failed Staging environment transformation An analytics engineer notices that staging data looks wrong. They ask their agent to check the details using `get_job_details`. The agent reports which specific nodes failed and why, letting them fix it immediately without opening any dashboards. ### Testing a new pipeline segment before full deployment A data engineer wants to test a small change on the Dev environment. They ask their agent to `trigger_job` for that specific job, ensuring they don't accidentally run it against Production or waste compute resources. ### Getting an overview of all possible environments A new team member needs to know where the data lives. They ask their agent to `list_environments`, which instantly returns a list of 'Dev,' 'Staging,' and 'Production' with status indicators, getting them up to speed in seconds. ## Benefits - Quickly check job status: Instead of navigating logs, you simply ask your agent to check the run progress. This saves minutes on every debugging session. - Targeted execution: You don't need full UI access to start a pipeline. By using tools like `trigger_job`, you can isolate and test specific transformations directly through conversation. - Full visibility: With one call, your agent runs `list_environments`, giving you an instant overview of every setup in your organization—a massive time saver for data leads. - Debug from chat: If a run fails, your agent doesn't just say 'fail.' It checks the status and helps retrieve job details, letting you debug complex failures instantly. - Better process control: You manage transformations via natural language. This keeps your entire workflow history centralized with your AI client, not scattered across multiple tabs. ## How It Works The bottom line is, your AI client handles all the API communication; you just talk to it like talking to a teammate. 1. Connect this MCP to your preferred AI client via Vinkius, providing the necessary API token credentials. 2. Tell your agent what you need done using plain English. For example: 'What is the status of the Production environment run?' 3. The agent executes the correct tool call against Coalesce and presents the structured data—like job IDs or progress percentages—back to you in conversation. ## Frequently Asked Questions **How does the Coalesce MCP help me monitor job status?** The Coalesce MCP lets your agent check pipeline progress instantly. You can ask for a current run status, and it tells you the percentage complete or if there was an error, saving you from manually checking dashboards. **Can I use Coalesce MCP to start new data pipelines?** Yes. You can trigger jobs and full pipeline runs on demand. This means when a test is needed or an urgent update hits, your agent starts the transformation for you without needing UI access. **Does Coalesce MCP work with my existing Snowflake setup?** Absolutely. Because it connects directly to the Coalesce platform built on Snowflake, it manages transformations and data pipelines exactly where your data lives, making everything cohesive. **What information does the Coalesce MCP give about environments?** It gives you a full picture. You can list all configured environments—Dev, Staging, Prod—and pull specific details for any one environment to verify its setup parameters. **What if I need to debug a failed run using Coalesce MCP?** You just ask your agent. It can check the job's history and retrieve detailed logs, pointing out exactly which step or node caused the failure. This cuts down debugging time from hours to minutes.