Picture this. You’re deep in a Claude conversation, your research is loaded, your prompts are dialed in, and then you need a visual. So you stop, open another tab, copy everything over, and rebuild your context somewhere else. That little break kills the momentum every single time.
I came across a post from an AI professional who lives in exactly that workflow, and the fix made me sit up. The author uses Claude as their default engine but kept getting yanked out of the conversation every time they wanted an image or a video. Now, they say, Runway runs directly inside Claude. No tab-switching. No context rebuild.
So what actually changed?
Runway launched its MCP, and the creator broke down what that means in plain terms. MCP is the connector between an AI assistant and the tools it reaches for. The Runway MCP links Claude to Runway, so you generate visuals from the same conversation you’re already having.
Here’s the part the original poster flagged that I think matters most: it works across Claude Chat, Claude Desktop, Claude Code, and other MCP compatible clients like Codex. So you’re not locked into one surface. You pick where you already work.
No API key, which is the nice surprise
According to this AI professional, you don’t need an API key to make it run. You just need a Runway account. The setup is short:
- Open a Runway account if you don’t have one
- Add Runway as a custom connector inside Claude
- Follow the setup steps
- Start generating visuals from your conversation
That’s it. The expert called it ready to go after those steps, and the low setup bar is a big reason this feels approachable instead of intimidating.
The context is the real win
This is the part where I got genuinely excited. The creator pointed out that you can bring your own context into the visuals. Your documents. A website. Your own research. Then you use that material to drive what gets generated.
So instead of describing an idea from scratch every time, you’re building visuals on top of work you already did. The post’s author also mentioned leaning on the Claude Skills and workflows they already use for writing, prompting, and research, then generating from there.
A couple of specifics the creator named:
- Seedance for generating videos
- GPT Image 2 for building storyboards
All of that without switching tools. You stay in one place, and the creative chain stays unbroken.
Why it matters: the friction of jumping between tools is one of those small taxes that quietly drains a creative session. Pulling generation into the same window where your context already lives is the kind of change that compounds over a workday.
Where this fits the bigger picture
MCP connectors are popping up fast, and this is a clear example of why people are paying attention. The pattern is simple: keep the AI assistant at the center, then plug specialized tools into it instead of scattering your work across ten tabs. If you create images or video as part of your job, this is a setup worth watching.
The mind behind the post was honest that it’s still early. They said it already feels like the start of a much smoother creative workflow, and they plan to share more after testing it further. I think that’s the right read. Early, but pointing somewhere good.
If you run your creative work through Claude, the full LinkedIn post is worth a look for the setup details and the creator’s first impressions. Check it out.