Grok

@Grok list all available models with their pricing and show me which ones support vision input

Here are the models available on your xAI account: **Text Models:** - grok-beta (128k context, $5/1M input tokens, $15/1M output) - grok-vision-beta (128k context, vision-capable, $5/1M input tokens, $15/1M output) **Embedding Models:** - grok-embedding-1 (8k context, $0.10/1M tokens) **Image Generation:** - grok-2-image-1 ($0.05 per image) The grok-vision-beta model supports multimodal input including images. All pricing is current as of your API key's access level.

This example shows the List Models tool, which is essential for understanding what capabilities your API key can access. Model availability and pricing can change; the MCP returns live data from xAI's API. Useful for cost estimation before running expensive operations like video generation or high-volume embedding jobs.

@Grok generate an image of a minimalist workspace with a laptop, coffee mug, and morning light through a window

Image generated successfully. The xAI image model produced a 1024×1024 rendering showing a clean desk with a silver laptop, white ceramic mug, and soft natural light casting shadows across the surface. The composition uses a muted color palette with warm highlights. **Usage:** 1 image generated, $0.05 charged to your account. The image is available for download and has been stored temporarily. Note that xAI image generation is synchronous—this request completed in approximately 8 seconds.

Demonstrates the Generate Image tool for visual content creation. Unlike video generation, image requests are synchronous and return immediately. Be aware that image generation costs accumulate quickly with multiple requests, and xAI's content policy applies—prompts violating guidelines will be rejected before generation starts.

@Grok start generating a 5-second video of ocean waves at sunset, then check the status and let me know when it's ready

Video generation started. Request ID: `vid_abc123def456` Initial status: **Pending** — xAI is processing your request. Video generation typically takes 2-5 minutes depending on queue depth. *[After polling with Get Video tool]* Status update: **Completed** — Your 5-second video is ready. The clip shows rhythmic ocean waves with orange and pink sunset reflections. Video resolution: 1280×720, format: MP4. **Usage:** 5 seconds generated, $0.50 charged to your account ($0.10/second). Video generation is asynchronous, so you'll need to poll for completion or set up a workflow to check back periodically.

Showcases the async video workflow using Generate Video and Get Video tools together. Video generation is significantly more expensive than images ($0.10/second) and requires polling to check completion status—the MCP doesn't automatically wait. Plan for 2-5 minute delays and handle pending states in your automation logic.

When Grok's embeddings beat a dedicated vector service for small teams

A 6-person support team wants to embed 2,000 help articles so agents can search by semantic similarity during live chats. Grok's Create Embeddings tool handles this in one API call with no infrastructure setup—no Pinecone account, no separate embedding pipeline. The trade-off: you're locked into xAI's embedding model and pricing, and there's no built-in vector store, so you'll still need a lightweight database like SQLite with pgvector or Supabase to hold the results. If your corpus grows past 10,000 documents or you need sub-100ms retrieval, a dedicated vector service becomes worth the setup cost. For teams under that threshold who already use Grok for chat, consolidating embeddings here cuts one vendor and one auth flow.

Grok's image and video tools for rapid campaign prototyping

A 3-person growth team needs 20 social-media variants and 5 short-form video concepts for a product launch in 48 hours. Grok's Generate Image and Generate Video tools let them prompt, iterate, and export without leaving the Switchy workspace or juggling Midjourney and Runway accounts. The video tool is asynchronous—you kick off generation, poll with Get Video, and retrieve the file when ready—so it fits batch workflows better than real-time editing. The catch: xAI's models are general-purpose, not fine-tuned for brand consistency, so expect to spend 10-15 minutes per asset on prompt refinement. If your brand guidelines are strict or you need frame-level control, dedicated creative tools still win. For fast, low-fidelity prototyping where speed trumps pixel-perfect output, Grok closes the loop in one MCP.

Using Grok's account and model tools to track xAI spend across projects

A 12-person engineering org runs Grok calls from three different projects—chatbot backend, internal docs search, and a prototype reasoning agent. The Get Current Account and List Models tools surface which API keys are hitting which models, plus per-model pricing metadata, so the team lead can attribute costs without parsing raw billing CSVs. Get API Key Info exposes ACL scopes, which matters when you want to lock a staging key to embeddings-only or block a deprecated key before it racks up charges. The limitation: these are read-only inspection tools, not cost-control levers—you can't set spend caps or auto-rotate keys from the MCP. If your monthly xAI bill is under $500, this visibility is enough to catch runaway usage. Above that threshold, you'll want xAI's dashboard or a third-party spend tracker.