i2v.ai

Generate images and videos across 12 image and 16 video models with a USD quote before every paid task.

Servidor MCP hospedado

npx add-mcp 'https://i2v.ai/api/mcp'

Instala no Claude Code, Codex, Cursor e outros

Documentação

i2v.ai MCP

Connect i2v.ai over MCP to find models, check the cost of your actual inputs, and generate images or videos from the client you already use.

Choose your MCP client

Run this command in the terminal where you use Codex CLI.

codex mcp add i2v --url https://i2v.ai/api/mcp

Sign in to i2v.ai in the browser when prompted. If you still need to authorize, run codex mcp login i2v.

Enable generation and uploads

To upload or generate, authorize again and enable write access on the i2v.ai consent screen.

codex mcp login i2v --scopes mcp:read,mcp:write

Verify your connection

Send this to your connected client to check the current image model catalog without submitting a task.

list_models

Call i2v.ai’s list_models with mediaType set to image. If there are more results, use offset to fetch the remaining pages until you reach total. Show the total count and list each model name and ID. Do not create a task.

Check balance and spend

Use i2v.ai get_account to show my available and reserved USD and granted scopes. Use get_usage for the last 30 UTC calendar days. Show charges, usage refunds and net spend for that window.

From a request to a result

Find a model

Filter by the task, reference media, and output requirements.

recommend_models

Check the quote

Validate your actual inputs and review the USD reservation before submitting.

prepare_generation

Submit and resume

Set your approved budget. Keep the task ID or original idempotency key for recovery.

create_task · recover_submission

Try a creative brief

Product image

Use i2v.ai to plan a studio product photo of a blue ceramic mug. Show the model, aspect ratio, and USD quote for the actual inputs. After I approve the budget, submit once with that cap and the model revision. Give me the task ID and the result when ready.

Image to video

Use my reference image as the first frame of a five-second vertical video. Find a compatible i2v.ai model, keep the subject consistent, and show me the settings and USD quote. Wait for my approval before creating a task.

Permissions, costs, and files

Which permissions should I grant?

OAuth starts with read access for discovery, quotes, and task lookups. Grant write access for uploads and submissions. An API key can read and submit tasks.

When is money reserved?

Connecting, listing models, and preparing a quote are free. A generation submission can reserve USD. Set maxCostUsd to cap the initial reservation, and use expectedRevision from the quote.

How do I use my own assets?

Use a public HTTPS URL or an upload supported by your client. Local stdio can upload or save files within I2V_WORKSPACE_DIR. Remote MCP returns result URLs.

What if the response is interrupted?

Look up the task ID with get_task. If the ID was lost, use recover_submission with the original credential and idempotency key. For OAuth recovery, set recoveryScope to connection when first submitting.

How do updates reach my client?

Hosted MCP tools are updated on i2v.ai. Refresh the tool list or reconnect if your client has cached it. Local stdio checks for versions when it connects. Run i2v-cli.mjs doctor to review the CLI, bundled Skill and server versions. Install updates explicitly and restart a local stdio host after replacing its script.