# Langflow Multi-agent MCP MCP > Langflow MCP lets your agent run complex AI workflows through natural conversation. You can manage entire projects, list all defined flows, execute specific multi-step chains, and trigger external webhooks—all from a single chat interface. ## Overview - **Category:** developer-tools - **Price:** Free - **Tags:** multi-agent, ai-orchestration, visual-programming, langflow, workflow-automation ## Description Running sophisticated, multi-agent systems used to mean jumping between dashboards or writing wrapper scripts just to test a simple change. Now you don't have to. This MCP connects your AI client directly to your Langflow instance, giving you full operational control over complex agent workflows via conversation. Need to check the logs after a run? Just ask. Want to update the logic for an old flow or create a new project folder? You can do it by talking to your agent. It’s essentially wrapping up all your workflow management into one conversational endpoint, making deep agentic work accessible right where you're already working. If you use Vinkius, this MCP gives you instant access to managing and running these complex AI pipelines without ever leaving your chat window. ## Tools ### create_flow Creates a brand new AI flow definition within Langflow. ### create_project Establishes a new container folder for organizing related projects. ### create_response Simulates an OpenAI-compatible response endpoint using a specific flow ID as the model. ### delete_file_v2 Removes a user file from the system storage. ### delete_flow Deletes an existing AI flow definition entirely. ### delete_project Wipes out all contents and definitions within a project folder. ### get_file_v2 Downloads the content of a specified user file. ### get_flow Retrieves the full configuration details for one specific flow using its ID. ### get_logs Fetches recent system logs related to workflow execution. ### get_monitor_messages Retrieves the complete chat history from a specific monitor session. ### get_monitor_traces Pulls detailed execution paths and span trees for debugging failures. ### get_monitor_transactions Retrieves logs detailing how different components interacted during a run. ### get_project Gets the current details and metadata for a designated project folder. ### list_files_v1 Lists files associated with a specific flow ID (version 1). ### list_files_v2 Lists all user-uploaded files in the system storage (version 2). ### list_flows Retrieves a list of every flow definition currently stored. ### list_projects Displays all active project folders and their IDs. ### list_users Lists details for all authenticated users in the system (requires superuser access). ### run_flow Executes a defined AI flow using either plain text or conversational input. ### run_workflow Runs a complex, long-running background workflow job (version 2 API). ### trigger_webhook Starts a flow run by simulating an external system webhook call. ### update_flow Modifies the parameters or configuration of an existing AI flow definition. ### update_project Updates metadata or descriptive information for a project folder. ### whoami Returns the profile and details of the currently authenticated user. ## Prompt Examples **Prompt:** ``` List all my available flows in Langflow. ``` **Response:** ``` I've retrieved your flows. You have 3 active flows: 'Customer Support Agent' (ID: flow-123), 'Data Extraction Pipeline' (ID: flow-456), and 'Market Analyzer' (ID: flow-789). ``` **Prompt:** ``` Run the 'Market Analyzer' flow with the input 'Analyze NVIDIA stock performance'. ``` **Response:** ``` Executing 'Market Analyzer'... The flow has completed. The analysis suggests a bullish trend for NVIDIA based on recent quarterly results and AI demand. ``` **Prompt:** ``` Show me all my project folders. ``` **Response:** ``` I found 2 projects: 'Production Agents' (ID: proj-001) and 'Experimental Lab' (ID: proj-002). ``` ## Capabilities ### Run Defined Workflows Execute entire multi-step AI processes using either text or conversational input. ### Manage Project Folders Create, read, and delete project folders to keep related agent workflows organized. ### Audit Agent Flows List all available flows or retrieve specific flow definitions by ID. ### Monitor Execution State Retrieve historical chat messages, execution traces, and component interaction logs for debugging. ### Trigger External Events Initiate a workflow run in response to an external system event via webhooks. ## Use Cases ### Debugging an agent failure The market analysis agent fails midway through. Instead of guessing what broke, you ask your agent to use `get_monitor_traces`. It immediately pulls the full execution path and span tree, showing exactly which component failed and why. ### Deploying a scheduled report A monthly compliance report needs to run. You don't write a script; you simply tell your agent to `run_workflow` using the designated background job ID, ensuring the process completes outside of real-time chat. ### Organizing experimental pipelines Your team builds three related prototypes. Instead of having them cluttering up the main list, you tell your agent to `list_projects` and then create a new folder using `create_project`, keeping all related flows contained. ### Handling real-time external data A ticket comes in via Zendesk. Instead of manually updating the CRM, you tell your agent to use `trigger_webhook` with the ticket ID. This starts the workflow that automatically updates all necessary records. ## Benefits - You don't have to leave your IDE or chat client. You can execute any flow using `run_flow` directly from your conversation, eliminating context switching between the AI platform and the terminal. - Keep projects clean. Use `list_projects` and `create_project` to group related agent workflows into distinct folders, making version management simple for product teams. - Debugging complex chains is easier than ever. Instead of digging through server logs, you can ask your agent to retrieve specific details using `get_monitor_traces` or `get_monitor_messages`. - Automate the whole system. If an external event happens—say, a new record hits a database—you don't need middleware. You just use `trigger_webhook` to fire off the workflow immediately. - Full lifecycle control: Need to fix a broken flow? Use `get_flow` to check the definition, and then `update_flow` to patch it up without rewriting code. ## How It Works The bottom line is you get full control over your complex AI agent environment, managed purely by chat commands. 1. Subscribe to this MCP and provide your Langflow Base URL and API Key. 2. Your agent connects the credentials, making all defined flows and project structures available through natural conversation. 3. You issue a command (e.g., 'Run the Market Analyzer flow with input X'), and the MCP executes the action and returns the result directly. ## Frequently Asked Questions **How do I see what flows are available using list_flows?** Use `list_flows` to retrieve a directory of all defined agent pipelines. This command gives you the IDs and names, so you know exactly which flow you need to execute. **What is the difference between run_flow and run_workflow?** `run_flow` handles typical, contained chat-based agent interactions. `run_workflow` is for running complex, long-running background jobs that need to complete regardless of your current connection. **Can I check the history using get_monitor_messages?** Yes. You use `get_monitor_messages` to pull the full chat transcript from a specific session. This is essential for reviewing what the agent actually told you during its run. **How do I start an external process?** You trigger external events using `trigger_webhook`. You just need to provide the webhook ID, and the MCP simulates receiving that event, kicking off a flow immediately. **I'm organizing many agents. How do I use `list_projects` and `create_project` to keep my workflows separate?** Use this MCP to containerize your agent work into projects. First, call `list_projects` to see existing containers. Then, run `create_project` when you need a new folder for related flows or experiments. **I ran a complex flow and it failed. How do I use `get_monitor_traces` to find the exact error?** The traces give you the full execution path, which is critical for debugging. Call `get_monitor_traces` to retrieve component interactions and span trees, showing exactly where and why the agent logic broke. **I need to download a user's data file linked to an agent flow. Should I use `list_files_v2` or `get_file_v2`?** Start by calling `list_files_v2` to see all available files for a specific flow ID. Once you have the correct file ID, run `get_file_v2` to actually download and retrieve the content. **Before I delete or update anything, how do I verify my user identity using `whoami`?** Use `whoami` to confirm your current authenticated credentials. This is key for verifying permissions before running sensitive actions like deleting a flow or modifying project metadata. **Can I run a flow using its name instead of a long UUID?** Yes! The `run_flow` tool accepts either the Flow ID or the Flow Name in the `flow_id` parameter, making it easy to trigger specific logic by name. **How do I see all the available projects and folders in my Langflow instance?** Use the `list_projects` tool. It will return a list of all projects (folders) which help organize your flows and components. **Is it possible to trigger a flow from an external webhook payload?** Absolutely. Use the `trigger_webhook` tool by providing the `flow_id` and the `data` JSON payload you want to send to the flow's entry point.