AI Video Atlas

What AI Video Atlas MCP server does

AI Video Atlas MCP server exposes the data behind AI Video Atlas through MCP and a REST API. It helps an agent turn a plain-language video idea into a sequence of shots, with a beat breakdown, a suggested model, a sourced prompt, and an estimated cost for each part.

The catalog contains prompts that creators have posted publicly, along with creator credit, the model used, estimated cost, and a link to the original material. Vendor pricing and model documentation are also included, so results can be checked against their stated sources rather than treated as unsupported estimates.

The service is read-only. It does not generate video, rewrite prompts through a hidden model, or modify Atlas data.

How it works

The hosted endpoint is available at https://mcp.aivideoatlas.com/mcp using streamable HTTP. It requires no authentication, account, API key, or cookie. MCP clients can call the tools directly, while applications that prefer HTTP can use the REST routes under https://mcp.aivideoatlas.com/v1/. An OpenAPI document is available at https://mcp.aivideoatlas.com/openapi.json.

Responses are assembled from Atlas data and include source links where applicable. The server also provides MCP resources such as atlas://guides/<model> and instructions that help agents select the relevant tool.

The public service permits about 60 requests per minute per client address. Requests above the limit receive HTTP 429 and a retry-after response value. Queries should not contain personal information.

Setup and configuration

For a hosted connection, add the MCP URL to a compatible client and choose no authentication. Claude, Cursor, Gemini CLI, VS Code, Windsurf, and other MCP clients can use the same endpoint. Claude Code can add it with an HTTP transport command, and Cursor can configure it in ~/.cursor/mcp.json.

A local deployment requires Node.js dependencies, a build, and Cloudflare Workers tooling. Run npm install, then npm run build; the build downloads the public Atlas export and embeds it in dist/worker.mjs. npx wrangler dev starts a local instance, and npx wrangler deploy deploys it to a Cloudflare account. A local data export can be supplied to the build script instead. No secrets are required, although deployment needs a Cloudflare account and an account ID configured in Wrangler or through CLOUDFLARE_ACCOUNT_ID.

Tools and capabilities

• plan_shots breaks a description into beats and selects sourced shots, models, prompts, and estimated costs.
• find_shot searches the library of creator-supplied AI video shots.
• best_prompt returns the closest full sourced prompt, optionally filtered for a target model.
• recommend_model compares model usage evidence, the lowest vendor-stated cost per second, and an estimated total.
• get_prices lists vendor-stated prices per second with source URLs and check dates; discontinued entries are omitted unless requested.
• get_workflow provides workflows and related tools or models for tasks such as lip-sync, upscaling, and character consistency.
• get_guide returns model-specific advice covering prompt structure, camera direction, modes, consistency, audio, limitations, and avoidances.
• whats_new finds shots and guides added within a selected number of days.

The AI Video Atlas MCP server is suitable for pre-production research, shot breakdowns, prompt discovery, model comparison, and budget estimation. It is not a rendering or generation service.