Server data from the Official MCP Registry
Plan AI video shots: model picks, sourced prompts, prompting guides and real $/s costs.
About
Plan AI video shots: model picks, sourced prompts, prompting guides and real $/s costs.
Remote endpoints: streamable-http: https://mcp.aivideoatlas.com/mcp
Security Report
The visible worker code is a read-only, keyless MCP/REST server that builds deterministic answers from inlined Atlas data, with no shell execution, eval, or outbound calls to user-supplied URLs, and its network access matches its purpose. The main concerns are the intentionally unauthenticated public endpoint (abuse is limited only by per-IP rate limiting), anonymized query-text logging, and third-party creator text being returned to LLM clients. The truncated source and the unreviewed build script leave some verification gaps. Supply chain analysis found 1 known vulnerability in dependencies (0 critical, 1 high severity).
3 files analyzed · 7 issues found
Security scores are indicators to help you make informed decisions, not guarantees. Always review permissions before connecting any MCP server.
Permissions Required
This plugin requests these system permissions. Most are normal for its category.
How to Install & Connect
Available as Local & Remote
This plugin can run on your machine or connect to a hosted endpoint. during install.
Documentation
View on GitHubFrom the project's GitHub README.
AI Video Atlas MCP server
Give your AI agent a video director. This is the open-source MCP server (and REST API) behind AI Video Atlas. It lets Claude, ChatGPT, Cursor, Gemini, Grok Bot or any MCP client plan AI video shots, pick models, find sourced prompts, and price a clip using vendor-stated $/second, all from Atlas data.
- Live endpoint:
https://mcp.aivideoatlas.com/mcp(streamable HTTP, no auth, free) - REST API:
https://mcp.aivideoatlas.com/v1/<tool>. The OpenAPI spec is athttps://mcp.aivideoatlas.com/openapi.json. - Agent guide: https://aivideoatlas.com/agents/
It's read-only and deterministic: no AI generation happens on the server, and every answer is built from Atlas's sourced data, with links back to the original creators and vendor pages.
Connect
Claude (web and desktop): open claude.ai → Settings → Connectors, click Add custom connector, name it AI Video Atlas and paste https://mcp.aivideoatlas.com/mcp. No sign-in is needed. On Team and Enterprise plans, an owner may need to add the connector first.
Claude Code:
claude mcp add --transport http ai-video-atlas https://mcp.aivideoatlas.com/mcp
ChatGPT: custom MCP connectors need developer mode, a beta for paid plans (Pro, Business, Enterprise, Edu). It isn't available on Free. Open Settings → Apps & Connectors → Advanced settings and turn on Developer mode. Then click Create, name it AI Video Atlas, paste the URL and set authentication to No authentication.
Building a custom GPT instead? Go to Configure → Actions → Import from URL and use https://mcp.aivideoatlas.com/openapi.json.
Cursor: use the one-click "Add to Cursor" button on https://aivideoatlas.com/agents/, or add this to ~/.cursor/mcp.json:
{"mcpServers":{"ai-video-atlas":{"url":"https://mcp.aivideoatlas.com/mcp"}}}
Gemini: Gemini CLI (free):
gemini mcp add --transport http ai-video-atlas https://mcp.aivideoatlas.com/mcp
In the Gemini app, custom connected apps are limited to Gemini Spark on personal Google accounts in the US (18+). If you have it, add a custom app with the URL above. Gems can't connect to outside tools.
Grok Bot: paste this into your Grok Bot chat:
Add this MCP server: name aivideoatlas, URL https://mcp.aivideoatlas.com/mcp (no key needed).
Grok Bot confirms before adding it, and the tools are available from your next message.
Any MCP client / API: use the URL above (streamable HTTP, no auth). For plain HTTP:
curl "https://mcp.aivideoatlas.com/v1/find_shot?query=drone+shot&limit=3"
VS Code, Windsurf and other MCP clients take the same URL.
Tools
| Tool | What it does |
|---|---|
plan_shots | Plan a multi-shot video from a plain description: splits it into beats and picks a sourced shot, model, prompt and estimated cost for each. Deterministic, built only from Atlas data. |
find_shot | Search the library of sourced AI video shots (real prompts creators posted, with the model, estimated cost and a link to the original). |
best_prompt | Get the full sourced prompt that best matches a shot, optionally for a target model, with creator credit and source. |
recommend_model | Recommend a model for a shot with evidence: how many sourced shots used each model, the cheapest vendor-stated $/s, and an estimated total. |
get_prices | Vendor-stated $/second prices with source URL and check date. Filter by family or host. Discontinued rows are excluded unless asked for. |
get_workflow | How-to workflows, tools and model guides grouped by job (lip-sync, upscaling, character consistency and more). |
get_guide | Per-model prompting guidance (structure, camera, modes, consistency, audio, what to avoid, limits), sourced from vendor docs plus proven Atlas examples. |
whats_new | The newest shots and guides added in the last N days. |
The server also exposes MCP resources (for example, atlas://guides/<model>) and server instructions that tell agents when to use each tool.
Rate limits
The public instance is free and keyless. It allows about 60 requests a minute per client (per IPv4 address, or per IPv6 /48), enforced by a Durable Object. A backup edge rule also applies. Over the limit, you get HTTP 429 with retry-after.
Privacy
- No accounts, keys or cookies.
- IP addresses are never stored or logged. They're used only in memory to apply the rate limit.
- Workers invocation logs are off. The only log line is an anonymous
atlas_queryevent: tool name, query text (lowercased and truncated), argument names, result count, latency and client name, if the client sends one. These are used to see which tools and topics are useful. - Daily aggregate counts (sessions, tool calls by tool, MCP vs REST, client names) are kept in a Durable Object and published at https://mcp.aivideoatlas.com/stats/daily. Counts only: no IPs, no query text.
- Don't put personal information in queries.
- Full policy: https://aivideoatlas.com/privacy/
Run your own copy
npm install
npm run build # downloads the public data export from https://aivideoatlas.com/agent-data.json and inlines it
npx wrangler dev # local
npx wrangler deploy # needs your own Cloudflare account
npm run build writes dist/worker.mjs. To build from a local export instead, run node scripts/build.mjs --data=path/to/agent-data.json, or place it at data/agent-data.json (gitignored).
In wrangler.toml, set account_id (or use CLOUDFLARE_ACCOUNT_ID) and, optionally, your own custom-domain route. The Durable Object binding LIMITER → RateLimiter (SQLite-backed, available on the free plan) is already declared. No secrets are required.
Data and licensing
The code is MIT-licensed (see LICENSE). The data is not in this repo: it's AI Video Atlas's public export, with credit to the original creators and vendor sources linked on every item. Please keep those credits when you show results.
Registry
server.json is the entry for the official MCP Registry (com.aivideoatlas/atlas).
Contact: hello@aivideoatlas.com
Reviews
No reviews yet
Be the first to review this server!
More Developer Tools MCP Servers
Fetch
Freeby Modelcontextprotocol · Developer Tools
Web content fetching and conversion for efficient LLM usage
Git
Freeby Modelcontextprotocol · Developer Tools
Read, search, and manipulate Git repositories programmatically
Worldmonitor
Freeby Koala73 · Developer Tools
Live markets, conflicts, country risk, chokepoints, energy, and China decision signals. 93 tools.
Toleno
Freeby Toleno · Developer Tools
Toleno Network MCP Server — Manage your Toleno mining account with Claude AI using natural language.
MCP Marketplace
Freeby mcp-marketplace · Developer Tools
Search and install MCP servers from inside your AI client.
mcp-creator-python
Freeby mcp-marketplace · Developer Tools
Create, build, and publish Python MCP servers to PyPI — conversationally.
