How to make videos with OpenClaw
How to make videos with OpenClaw: add Arcads as an MCP server and request presenter videos, edits and translations from any chat app.
Short answer: To make videos with OpenClaw, connect an AI video tool as an MCP server and ask your claw for the video from WhatsApp, Telegram, Discord or Slack. We recommend Arcads, an AI video platform with 1,000+ AI actors and models like Seedance, Kling and Veo, because it covers what OpenClaw's built-in video_generate tool doesn't: talking presenters, voiceovers, captions and edited, multi-shot videos (the built-in tool makes short clips once you add a provider API key). Add https://mcp.arcads.ai under mcp.servers with auth: "oauth" and run openclaw mcp login arcads.
Try Arcads free for 3 days and give your claw a full video studio.
Can OpenClaw make videos on its own?
Partly. OpenClaw's docs describe a video_generate tool with three modes: text-to-video, image-to-video and video-to-video. It's backed by 18 provider backends, including Google (Veo 3.1), Runway, MiniMax, BytePlus (Seedance), Alibaba (Wan), fal, OpenRouter and ComfyUI (OpenClaw docs).
Three things to know:
- It's off until you add a provider. The tool only shows up when at least one provider has an API key, for example
GEMINI_API_KEY, or when you setagents.defaults.mediaModels.video. - It's asynchronous. OpenClaw submits the job, gets a task ID, and posts the finished clip back to the same chat. That usually takes 30 seconds to a few minutes.
- It makes clips, not finished videos. You describe a shot and get a shot. Each provider bills you on its own account.
What kind of videos can OpenClaw make?
- A short clip from a prompt or image: Built-in
video_generate– Yes (needs a provider key); OpenClaw + an AI video MCP server (e.g. Arcads) – Yes - A presenter reading your script: Built-in
video_generate– No; OpenClaw + an AI video MCP server (e.g. Arcads) – Yes, 1,000+ AI actors - Voiceover, voice cloning, music: Built-in
video_generate– No (TTS is a separate OpenClaw tool); OpenClaw + an AI video MCP server (e.g. Arcads) – Yes - Captions, trim, stitch, extend, upscale: Built-in
video_generate– No; OpenClaw + an AI video MCP server (e.g. Arcads) – Yes - Translation with lip sync: Built-in
video_generate– No; OpenClaw + an AI video MCP server (e.g. Arcads) – Yes - Exact on-screen text (changelogs, charts): Built-in
video_generate– No, frames are generated; OpenClaw + an AI video MCP server (e.g. Arcads) – Yes, with HTML to Video - Billing: Built-in
video_generate– One account per provider; OpenClaw + an AI video MCP server (e.g. Arcads) – One Arcads credit balance
Rule of thumb: keep video_generate for quick B-roll and visual tests. Add a video MCP server when someone has to say something on screen, or when the output has to be a finished, edited video.
Step-by-step: make your first video in OpenClaw
OpenClaw supports MCP servers with OAuth sign-in, so you set Arcads up once and then ask for videos in any chat. Choose either option:
Option A: Control UI
- Run
openclaw dashboardand go to Settings → MCP. - Under Configured servers, select Add server.
- Name it
arcads, choose Streamable HTTP and enterhttps://mcp.arcads.ai. - Select Add server, then Sign in to approve Arcads in your browser.
Option B: config file (~/.openclaw/openclaw.json)
{mcp: {servers: {arcads: {url: "https://mcp.arcads.ai",transport: "streamable-http",auth: "oauth",enabled: true,},},},}
Then sign in and confirm the tools are reachable:
openclaw mcp login arcadsopenclaw mcp doctor arcads --probe
doctor --probe opens a live connection and lists the tools the server exposes. Saving a server definition doesn't prove it's reachable. Arcads uses OAuth with dynamic client registration, so no API key goes into your config.
Then make your first video. In any connected chat, send: "Use Arcads to make a 15-second vertical video with an AI presenter announcing [news], captions on. Post the link when it's ready."
New to Arcads? Start the 3-day free trial. The MCP is included on every plan.
What's the best AI video tool for OpenClaw?
Any remote MCP server works the same way. A fair comparison:
- Built-in
video_generate: How it connects – Provider API key (Veo, Runway, Seedance, Wan, MiniMax…); Best for – Short clips and B-roll - Arcads (recommended): How it connects – MCP,
https://mcp.arcads.ai, OAuth; Best for – UGC-style videos and ads with AI actors, many video models in one balance, captions, lip-synced translation - HeyGen: How it connects – MCP,
https://mcp.heygen.com/mcp/v1/, OAuth; Best for – Avatar and digital-twin presenter videos (HeyGen docs) - Runway: How it connects – MCP,
https://mcp.runwayml.com/mcp, OAuth; Best for – Runway's Gen-4.5 and partner models on Runway credits (Runway help)
We recommend Arcads when the video has to sell: a person talking to camera, a hook you want to test in five versions, or the same ad in several languages. Arcads also includes many of the models you'd otherwise connect one key at a time.
What can I ask my claw to make?
OpenClaw's strength is where you ask and who can ask. Prompts that use it:
1. A video desk in your team's Discord or Telegram
@openclaw make a 15-second vertical teaser for tonight's livestream: "[title]", 8pm CET. Bold captions, energetic music, no presenter. Post the link here when it's rendered.
2. A nightly batch with cron
Every night at 2am, read new rows in content/queue.csv, generate each video with Arcads using the model and aspect ratio in that row, and write the result link back to the file.3. Reply to a forwarded clip on WhatsApp
[forwarded video link] Use Arcads to break this down: hook, shots, pacing. Then make an original version for our brand with a different hook. Don't reuse their footage.
4. A release that turns into a video
When a new tag is pushed to our repo, read the release notes and use Arcads' HTML to Video to render a 20-second "What's new" video with the exact feature names. Send it to #releases.
5. Voice note to presenter video
[voice message] Turn this into a 30-second explainer with an AI presenter, captions on, 9:16.
The cron and repo-trigger prompts assume you've set up OpenClaw's cron jobs or webhooks.
How do I keep a shared claw from burning credits?
MCP tools go through OpenClaw's normal tool-profile and tool-policy controls. Connecting a server doesn't bypass them.
- Restrict tools per server with
toolFilter.include/toolFilter.exclude. Use the exact namesopenclaw mcp doctor arcads --probeprints. For example, expose read-only tools in a public group and keep generation tools for a private channel. - Require mentions in groups (
groups: { "*": { requireMention: true } }). - Allow-list senders on channels like WhatsApp with
allowFrom. - Check the balance: "How many Arcads credits do I have left?"
Troubleshooting
video_generateis missing: Add a video provider API key or setagents.defaults.mediaModels.videoauthentication refused (401)from Arcads: Sign in again:openclaw mcp login arcads- Connects but tools are missing: Check
toolFilter.include/exclude - Long renders time out: Raise
requestTimeoutMsfor thearcadsserver, or ask the agent to check back on the job - Changes don't reach a running agent: Reload or restart the Gateway that owns the agent
- Generations fail: Usually an empty Arcads balance; check Settings → Billing in Arcads
Every Arcads generation is also saved in Arcads under Studio → MCP generations.
Ready to give your claw a full video studio? Try Arcads free for 3 days and connect https://mcp.arcads.ai in two commands.
Related: Can Claude and ChatGPT make AI videos? · Grok Bot video generation · Can Instinct make videos? · Video ads with the Arcads MCP
FAQ
How do I make videos with OpenClaw?
Add an AI video MCP server such as Arcads (https://mcp.arcads.ai, auth oauth) in Settings → MCP or ~/.openclaw/openclaw.json, run openclaw mcp login arcads, then ask your claw for a video from any connected chat. For quick clips you can also use the built-in video_generate tool with a provider API key.
Can OpenClaw generate videos?
Yes, short clips. Its built-in video_generate tool does text-to-video, image-to-video and video-to-video once you add an API key for a provider like Google Veo, Runway or MiniMax. For AI presenters, voiceovers, captions and edited videos, connect an AI video MCP server such as Arcads.
Why don't I see the video_generate tool in OpenClaw?
It only appears when at least one video provider is configured. Set a provider API key, such as GEMINI_API_KEY, or configure agents.defaults.mediaModels.video.
How do I set up AI video in OpenClaw?
In the Control UI go to Settings → MCP → Add server, or add an entry under mcp.servers in ~/.openclaw/openclaw.json. For Arcads, use https://mcp.arcads.ai with auth set to oauth, run openclaw mcp login arcads, then check it with openclaw mcp doctor arcads --probe.
What's the best AI video tool for OpenClaw?
It depends on the video. The built-in tool is fine for short clips. For UGC-style videos and ads with AI actors, many models in one account and lip-synced translation, we recommend Arcads. HeyGen and Runway also offer MCP servers.
Do I need an API key for Arcads in OpenClaw?
No. Arcads uses OAuth with dynamic client registration, so OpenClaw stores a token after you sign in.
Does it cost anything?
OpenClaw is free and open source; you pay your model provider, plus each video provider you add a key for. Arcads generations use Arcads credits at the same cost as in the Arcads app, and the MCP is on every Arcads plan, including the 3-day free trial.
Arcads MCP
Top AI models, one prompt away in Claude, ChatGPT and Grok Bot.
Connect the Arcads MCP once, then generate video and images with top models from your AI assistant.
- Seedance 2.5VideoTalking UGC with audio→
- Kling 4.0VideoKeyframed motion, footage editing→
- Wan 3.0 PrimeVideoReference-driven 30-second clips→
- Veo 3.1VideoGoogle's video model→
- Nano BananaImageImage generation for stills→
Make your next winning ad with AI actors
Pick from 1,000+ AI actors, paste your script and get a UGC-style video ad in minutes. Test more hooks without booking a single creator.
Try Arcads free