Krea Comes to Grok Bot: Image, Video and Billing Setup Explained

Krea announced its Grok Bot integration on October 23, making image and video generation available from the assistant. The official launch points to Krea's MCP integration page. This is a new way to access Krea's creative tools, rather than an announcement of a new underlying image or video model.

For a founder or designer, the appeal is straightforward: the conversation that develops a brief can also produce the first visual assets. The practical questions come next. Which model should do the work, where do the assets belong, and which balance pays for the iterations?

How to connect Krea to Grok Bot

Krea's setup instructions say to open Plugins in the Grok Bot sidebar. On mobile, open the avatar menu and then Plugins. Search for Krea, add the plugin, choose Authorize and complete the Krea sign-in in the browser.

The integration uses a Krea account rather than requiring an API key for the ordinary sign-in flow. Krea describes support for image generation, video, enhancement, trained styles and creative workflows. Larger jobs run in the background and return when ready.

I would begin with one inexpensive, clearly specified output rather than an open-ended campaign request. Give the subject, aspect ratio, intended placement and any non-negotiable details. Review that result before asking the agent to generate a batch. It is easier to correct a misunderstood brief before it becomes twenty variations.

Workspace choice determines billing

The MCP developer documentation says OAuth-connected work uses compute units from the workspace selected during consent. That choice also controls where uploaded assets are stored. To switch the bound workspace later, disconnect and reconnect.

An API-token connection generally uses the workspace's separate API balance instead. Agent-session tools are an exception and use compute units even with a token. These are different billing paths; signing in through the plugin should not be confused with adding a prepaid API balance.

This matters for a shared studio or agency account. I would decide the workspace before supplying client references, then confirm where the first generated asset appears. Keeping a project's inputs, outputs and spending together makes it easier to review what happened and avoid mixing personal experiments with client work.

Ask for a quote before a batch

Krea documents a preflight option for generation and node-app calls. Setting it to true returns a cost quote, available balance and affordability result without submitting a job. The response also indicates whether the price is estimated. Model discovery and schema inspection help the agent choose valid inputs before generation.

The pricing page lists a Free tier with 100 compute units per day and selected model access, alongside paid tiers with different allowances. That is not a promise that every video model or every Grok Bot workflow is free.

For a production test, I would ask for a quote, approve one representative output and measure the cost of reaching an acceptable result, including revisions. A low price for the first attempt can be misleading if the brief needs many retries.

The useful change is shorter distance between planning and creating. Keep model selection, spending and asset review visible in that same conversation, and the plugin becomes easier to use deliberately rather than as an unlimited generation button.

I'm Paras. I work at the intersection of AI, engineering and distribution.

I help AI founders turn technical launches and product updates into clear stories that developers and founders read and share.

Book a 30-minute call.