# Conductor MCP for AI Agents MCP

> Conductor (Netflix OSS) MCP gives your AI client control over complex, multi-step business processes. It lets you define, execute, monitor, and manage entire workflows for microservices directly through natural conversation. Stop clicking around dashboards; start orchestrating critical systems from your agent.

## Overview
- **Category:** industry-titans
- **Price:** Free
- **Tags:** workflow-orchestration, microservices, netflix-conductor, automation, process-management

## Description

Running distributed systems means managing dependencies that change constantly. This MCP connects your AI client to a powerful workflow engine, letting you treat complex process orchestration like simple chat commands. Instead of jumping between dedicated UIs to check task status or define new processes, you tell your agent what needs doing. You can list existing workflows and their versions, validate definitions before deployment, start new runs asynchronously, or even pause an entire sequence if something breaks down. If managing distributed tasks feels overwhelming, remember that Vinkius hosts thousands of MCPs, putting all enterprise systems into one place for your AI client to manage.

## Tools

### add_task_log
Adds a specific log entry to track activity within any given task step.

### bulk_pause
Pauses multiple workflows simultaneously, stopping them from executing further.

### bulk_remove
Deletes several workflow definitions or instances in a single operation.

### bulk_restart
Forces multiple workflows to restart their execution from the beginning.

### bulk_resume
Unpauses several workflows, allowing them to continue processing where they left off.

### bulk_retry
Attempts to re-run a selection of failed or stalled workflows in batches.

### bulk_search
Searches for workflow definitions using a list of specific IDs.

### bulk_terminate
Stops and removes multiple running or defined workflows at once.

### create_event_handler
Sets up a new event handler to listen for specific system events.

### create_task_definitions
Creates one or more reusable definitions for microservice tasks.

### create_workflow_definition
Defines a brand new, complex workflow structure.

### delete_event_handler
Removes an existing event handler from the system.

### delete_task_definition
Deletes a reusable task definition that is no longer needed.

### delete_workflow_definition
Permanently removes an entire workflow blueprint from the system.

### execute_workflow
Runs a full workflow execution immediately and waits for the result (synchronous).

### get_all_queues
Retrieves the number of pending tasks waiting across all defined queues.

### get_correlated_workflows
Finds workflows that are linked together using a common correlation ID.

### get_event_handlers
Lists all active event handlers currently registered in the system.

### get_queue_size
Checks how many tasks are waiting for a specific task type to run.

### get_running_workflows
Retrieves the IDs of all workflows currently in an active execution state, grouped by their type.

### get_task_definition
Fetches the details for a specific task definition using its name.

### get_task_definitions
Retrieves a list of all available and defined microservice tasks.

### get_task_logs
Fetches the detailed historical logs for a particular task execution.

### get_workflow_definition
Retrieves the full structure and details of one specific workflow definition by name.

### get_workflow_definitions
Gets a comprehensive list of all defined workflows available for execution.

### get_workflow_names_and_versions
Lists every workflow name and all versions that have been created for it.

### get_workflow_tasks
Shows the sequence of tasks required for a given running workflow instance.

### get_workflow
Retrieves the complete status and history of a single, specific workflow run by its ID.

### pause_workflow
Temporarily stops a specified running or defined workflow instance.

### poll_batch_tasks
Checks the status of multiple queued tasks over time using long polling.

### poll_task
Checks the current status of a single, specific task execution over time.

### remove_workflow
Permanently deletes a workflow from the system registry.

### requeue_tasks
Forces pending tasks back into the queue to be processed again.

### rerun_workflow
Re-runs a workflow starting from a specific task that failed or needs revalidation.

### restart_workflow
Resets and restarts an entire workflow execution back to its initial step.

### resume_workflow
Continues a workflow that was previously paused, allowing it to complete naturally.

### retry_workflow
Attempts to re-run only the most recently failed task in a sequence.

### search_workflows
Searches for workflows using basic criteria and returns a summary of the results.

### search_workflows_v2
Performs a detailed search across all workflows, returning full object details.

### skip_task
Bypasses a specific task step in an active workflow instance.

### start_workflow
Triggers a new workflow execution without waiting for immediate results (asynchronous).

### terminate_workflow
Immediately stops and cancels a running workflow instance.

### update_event_handler
Modifies the rules or targets of an existing event handler.

### update_task_by_ref
Updates a task using its internal reference name for precise control.

### update_task_definition
Modifies the operational details of a specific microservice task definition.

### update_task
Updates the result data for a completed task step.

### update_task_v2
Updates a task and automatically polls to find the next available task in the sequence.

### update_workflow_definitions
Creates or updates multiple workflow definitions simultaneously (batch management).

### validate_workflow_definition
Checks a workflow's JSON structure for errors without actually saving it to the system.

## Prompt Examples

**Prompt:** 
```
Can you list all available workflows and tell me which ones need updating?
```

**Response:** 
```
**Workflow Definitions Found:**
*   `UserOnboarding`: v1.2 (Needs review)
*   `PaymentProcessing`: v3.0 (Last updated 4 hours ago)
*   `InventoryCheck`: v5.1 (Active/Stable)

We found three primary workflows. Version `user_onboarding:v1.2` is marked for review because the 'billing' task has been deprecated in favor of a new service. I recommend checking that definition.
```

**Prompt:** 
```
I need to check the logs for the payment processing workflow that ran yesterday.
```

**Response:** 
```
Here are the key details for your request:

*   **Workflow ID:** `8f2b3c4d-5e6f...`
*   **Status:** Completed Successfully
*   **Last Task Run:** `CreditCardValidation`
*   **Log Snippet:** The task completed in 1.2s. Key output: *Transaction approved for $99.99.*

(Full log retrieval is available upon request.)
```

**Prompt:** 
```
Can we pause all the user onboarding workflows that are currently running?
```

**Response:** 
```
**Action:** Pausing Workflows
**Target:** All active `UserOnboarding` instances.

✅ Successfully paused 14 workflow instances.
*   The process will stop immediately at its current step.
*   You can resume them later by asking to 'resume the onboarding workflows'.
```

## Capabilities

### Manage Workflow Definitions
List, get, create, update, and validate the structure of entire workflow definitions before they run.

### Execute Workflows on Demand
Trigger new instances of any defined workflow immediately, or start a specific task within a running process.

### Monitor Running Processes
Get the current state of active workflows, check queue depth for specific tasks, and retrieve detailed logs for individual steps.

### Control Workflow State
Pause, resume, retry failed tasks, bulk restart, or terminate entire workflow sequences at runtime.

### Batch Operations
Perform large-scale actions like updating definitions or managing multiple workflows with a single command.

## Use Cases

### Onboarding a New Client Account
A new user signs up, triggering an onboarding workflow. The agent uses 'start_workflow' to kick off the process and monitors it with 'get_running_workflows', ensuring that profile creation, email verification, and initial billing setup all complete successfully.

### Handling a Failed Payment Batch
A massive batch of payments fails due to an API timeout. Instead of manually checking each record, the agent uses 'bulk_retry' across 50 different transaction workflows and then checks the results with 'get_task_logs'.

### Updating Core Business Logic
The billing department changes how tax is calculated. A developer uses 'validate_workflow_definition' to test the new JSON structure before committing, then uses 'update_workflow_definitions' to deploy the change across all related services.

### Responding to a System Incident
A critical queue backs up. The architect asks the agent to check the depth with 'get_queue_size', then uses 'requeue_tasks' and 'bulk_resume' to clear the backlog and resume normal operations.

## Benefits

- Instead of manually listing every workflow version, use 'get_workflow_names_and_versions' to see all blueprints across your system at a glance.
- Need to troubleshoot? Use 'get_task_logs' or check the state with 'get_workflow' to instantly understand exactly where and why a process failed.
- Avoid manual deployments. With 'update_workflow_definitions', you can push changes to dozens of definitions in one go, instead of creating them individually.
- When debugging a failure, don't restart everything. Use 'rerun_workflow' or 'retry_workflow' to target only the failed step for surgical fixes.
- Save time managing state with bulk operations like 'bulk_pause' and 'bulk_resume', controlling entire sets of processes without touching a dashboard.

## How It Works

The bottom line is, it takes complex system operations and turns them into simple conversational commands.

1. Connect your AI client to this MCP and provide the Conductor Server URL.
2. Ask your agent to perform a high-level task, like 'check the status of user onboarding workflows.'
3. The MCP translates that request into specific function calls, retrieving the data or executing the necessary actions for you.

## Frequently Asked Questions

**How does Conductor (Netflix OSS) MCP help me troubleshoot a broken workflow?**
It gives you deep visibility into running processes. Instead of guessing, you can ask for the current state using 'get_workflow' or drill down with 'get_task_logs' to pinpoint exactly which step failed and why.

**Can I run a new workflow without having to manually trigger it?**
Yes. You can instruct your agent to use the 'start_workflow' function, kicking off complex processes asynchronously so you don't have to wait for immediate results.

**Is Conductor (Netflix OSS) MCP good for updating multiple workflows at once?**
It's excellent for bulk changes. Use 'update_workflow_definitions' or 'bulk_terminate' when you need to apply the same change, like version bumping or stopping old processes, across dozens of definitions quickly.

**What if a task fails midway through? Can I fix it?**
Absolutely. You don't have to restart everything. Use 'rerun_workflow' or 'retry_workflow' to target only the failed step, saving you time and computational resources.

**How do I ensure my new workflow definitions are correct before deploying them?**
You can use 'validate_workflow_definition'. This checks your JSON structure for errors without actually saving or changing anything in the live system, making deployments much safer.