# Runway ML MCP > Runway ML MCP Server gives your AI agent access to advanced video tools. Generate cinematic clips from text using `text_to_video`, animate still images with `image_to_video`, or blend footage between two key frames using `interpolate`. It also lets you manage costs and check job status with tools like `get_task` and `cancel_task`. Stop leaving video production to separate web portals. ## Overview - **Category:** ai-frontier - **Price:** Free - **Tags:** video-generation, generative-ai, media-production, creative-tools, animation, model-inference ## Description Runway ML gives your AI agent direct access to high-end video tools. You'll generate, animate, and refine complex video clips—all without jumping between a bunch of different web portals. ### Generating Video from Scratch When you need a totally new scene, use **`text_to_video`**. This tool synthesizes an imaginative video clip using only your text prompt, giving you control over the model and how long it runs—you can specify 5 or 10 seconds. For quick cinematic sequences, call either **`gen3_turbo`** for fast 5-second clips or opt for **`gen4_turbo`** when you need high-quality footage from the advanced Gen-4 Turbo model. ### Animating and Blending Footage You can bring static pictures to life. Use **`image_to_video`** by providing a source URL, a model choice, and a duration; this animates that picture into a dynamic clip. If you need more control over the movement—say, guiding exactly how the camera moves or how the subject shifts—you'll use **`image_text_to_video`**. This tool takes both your reference image *and* detailed text prompts to guide the specific motion. For super smooth transitions between two distinct keyframes, run **`interpolate`**. It mathematically blends those two images together, creating continuous, flowing motion footage. ### Managing Your Jobs and Credits Video rendering costs credits, so you gotta keep track of what's running. Use **`list_tasks`** to pull up a list of your most recent generation tasks, helping you see the history and current status. To check on one specific job, call **`get_task`**; this tells you if the render succeeded and gives you the final output URL if it did. If things go sideways or you change your mind mid-render, use **`cancel_task`** to immediately stop the video job, saving your account credits. You can also check what kind of resources you've got left by running **`get_organization`**, which retrieves all your account details and billing info. ## Tools ### cancel_task Stops a video render job immediately, saving your account credits. This action is irreversible. ### gen3_turbo Generates quick 5-second videos using the Gen-3 Alpha Turbo model based on prompts. ### gen4_turbo Creates high-quality video footage using the advanced Gen-4 Turbo model. ### get_organization Retrieves your account details, including available credits and organizational billing information. ### get_task Checks the status of a specific video generation job and provides the final output URL if it succeeded. ### image_text_to_video Generates video by combining both an input image reference and a detailed text prompt to control motion. ### image_to_video Animates a static source image into a video clip, allowing you to specify the model and duration. ### interpolate Creates smooth motion footage by mathematically blending two distinct keyframe images together. ### list_tasks Retrieves a list of your most recent video generation tasks, helping you track history and status. ### text_to_video Generates a video clip from scratch using only a text prompt, giving you control over the model and duration (5 or 10 seconds). ## Prompt Examples **Prompt:** ``` Create a 5 second cinematic video showing a sunset over an alien planet using Runway Gen-3 Turbo. ``` **Response:** ``` Your Runway generation task was successfully submitted to the Gen-3 Turbo model. The task ID is rw_task_1234. You can ask me to poll this task in a moment to review the final video output. ``` **Prompt:** ``` Take this reference image URL and animate it with Gen-3 Turbo to make the camera slowly pan backwards. ``` **Response:** ``` Image to video request initiated using the `image_text_to_video` tool. Your task is now processing and I got the ID 'rw_task_9876'. The animation will be ready shortly. ``` **Prompt:** ``` List all my ongoing tasks on Runway to see if the video has finished rendering. ``` **Response:** ``` I've scanned your recent 10 tasks. I see that task 'rw_task_1234' just SUCCEEDED! Here is the MP4 link to your generated video. ``` ## Capabilities ### Generate video from text prompts You run the `text_to_video` tool and receive a task ID for synthesizing an entirely new, imaginative video scene. ### Create short cinematic clips (Gen-3/Gen-4) You call either `gen3_turbo` or `gen4_turbo` to quickly generate high-quality, model-specific video sequences. ### Animate a still image You use `image_to_video`, specifying a source URL and duration, which animates the static picture into a dynamic clip. ### Directly animate an image with motion prompts The `image_text_to_video` tool takes both a reference image and a text prompt to guide exactly how the camera or subject should move. ### Blend two images into video You run `interpolate`, which creates smooth, continuous motion footage between two distinct keyframe images. ### Track and manage generation jobs You use tools like `list_tasks` or `get_task` to check if a render succeeded and retrieve the final output URL. ## Use Cases ### Need to storyboard a commercial scene fast A director needs five different angles of a futuristic city. They prompt their agent: 'Generate 5 clips using `text_to_video` with a cyberpunk theme.' The agent runs the tool multiple times, giving them five distinct task IDs they can review instantly, letting them prototype faster than ever. ### Need to make an existing photo look cinematic A user has a great reference photo but needs it animated. They ask their agent to run `image_text_to_video` using the URL and a prompt like 'Slow pan backwards.' The tool handles both the image input and the motion direction, delivering a ready-to-use clip. ### Tracking expensive video renders A producer runs three large videos. They don't know if they finished or how many credits were spent. Instead of logging into the website, they use `list_tasks` to see all history and then run `get_organization` to check their remaining credit balance. ### Creating a seamless visual transition A video editor has two key pieces of footage but needs them connected by a smooth motion bridge. They tell the agent to use the `interpolate` tool, providing the URLs for both images. The server outputs the perfect transitional clip. ## Benefits - Stop context switching. You generate and manage videos—from `text_to_video` prompts to animating images with `image_to_video`—all without leaving your agent interface. - Control costs immediately. If a render job is going over budget or off track, use `cancel_task`. This saves you credits faster than checking the portal dashboard. - Go beyond simple generation. Need complex motion? Run `interpolate` to create smooth transitions between two keyframes; it's much better than just cutting clips together. - Maintain full visibility into your work. Use `list_tasks` and then `get_task` to check the status of a render, confirming if it succeeded and where the final video asset is located. - Flexibility in creation models. You get dedicated tools like `gen3_turbo` for quick 5-second tests and `gen4_turbo` for your high-stakes, polished sequences. ## How It Works The bottom line is: You talk to your AI agent, it calls the right video tool, and you get a trackable task ID that leads to your finished footage. 1. First, you enable the Runway ML integration and provide your secure Developer Portal token. This connects your AI agent to the video generation API. 2. Next, you prompt your AI client: 'Generate a 10-second clip of X using `gen4_turbo`.' The agent then calls the specific tool and initiates the rendering task. 3. Finally, you ask the agent to check status (e.g., 'What's the status for my last job?'). It uses `get_task` to report success/failure and provides the final MP4 link. ## Frequently Asked Questions **How do I check if my generated video with the `text_to_video` tool is ready?** You use the `get_task` tool, providing your task ID. The response will confirm status (like SUCCEEDED) and give you a direct MP4 link to download the finished clip. **Is there a way to stop a video job if I change my mind?** Yes, use `cancel_task`. Just provide the task ID for the running job. This immediately shuts down the rendering process and saves your credits. **What's the difference between `image_to_video` and `image_text_to_video`?** `image_to_video` animates a static image by giving it a duration and model. `image_text_to_video` is more advanced; you give it an image AND a directional prompt to control exactly how the camera or subject moves. **Can I check my video generation credits before running a big batch?** Absolutely. Run `get_organization`. This tool pulls up your current account details, including how many credits you have left and your billing status. **What credentials do I need to connect my AI client before running a task with `gen3_turbo`?** You must secure an API key from your Runway Developer Portal. This token authenticates your calls and ensures the server tracks usage against your account. Never share this key; it's required for all generation tools. **If a video job fails or encounters an error using `text_to_video`, how do I check the details with `get_task`?** The `get_task` tool provides detailed status codes beyond just 'SUCCEEDED' or 'FAILED'. Check the output payload for specific error messages to diagnose why the render stalled, such as invalid prompts or resource limits. **Are there rate limits when I generate multiple videos in quick succession using `gen4_turbo`?** Yes, Runway enforces usage tiers and API call limits. If you exceed your allocated quota, the server will return a specific HTTP 429 error code. You'll need to wait for the reset window or upgrade your subscription. **What image file formats does `interpolate` accept when creating smooth motion between two keyframes?** `interpolate` requires both source images (keyframes) to be provided in common high-resolution formats like PNG or JPG. The system processes these inputs directly, so ensure they are correctly sized for the desired output resolution. **How can I preview the output without fully committing my prompt?** Video generation actions using the Runway engines are directly invoked through asynchronous background tasks. Instead of an exact real-time preview, you can start small 5-second generations or iterate closely on your prompt texts until they produce the final aesthetic you want. **Does this integration process video rendering in real-time?** No. Video rendering is asynchronous. When you submit a prompt or an image, the AI will dispatch a background task to Runway's servers and receive a 'task_id'. You will need to ask the AI to verify the status of this task later, usually after a few minutes, to fetch the final video URL. **Can I use Gen-4 models for free?** No. Runway requires credits for all generations, and Gen-4 Turbo consumes significantly more credits than earlier models or standard actions. You must maintain sufficient active credits or a paid subscription on your Runway account for the AI to successfully execute the rendering jobs.