# Axiom Lab MCP Server & Claude Skills: Generate Commercial Videos in Claude Code & Cursor

> Connect the remote Axiom Lab MCP server (https://axiomlab.online/mcp) to Claude Code, Cursor, or your AI agent to turn local PDFs, specs, and photos into QA-checked commercial MP4 videos.

Canonical URL: https://axiomlab.online/blog/axiom-mcp-server-claude-code-cursor-video-guide
Published: 2026-10-19 | Updated: 2026-10-24 | Author: Axiom Lab

Most video MCP servers only wrap raw 5-second b-roll generators. The remote Axiom Lab MCP server (https://axiomlab.online/mcp) gives Claude Code, Cursor, and autonomous agents an end-to-end commercial video production pipeline: upload files, review a structured plan, confirm length, and download a QA-checked 1080p film.

## Why Developers Need an End-to-End Video Production MCP (Not Just Raw B-Roll)

If you build with **Claude Code**, **Cursor**, or custom AI agents, you have probably noticed that most "video generation MCP servers" only wrap raw diffusion endpoints (returning a 4-to-5-second silent clip). An agent cannot hand a 5-second silent clip to a founder or product team as a finished deliverable.

The **[Axiom Lab remote MCP server (`https://axiomlab.online/mcp`)](/developers/cli-mcp)** and **[Claude Skills (`/developers/skills`)](/developers/skills)** expose the full Axiom Lab production pipeline over standard HTTP MCP:

* **Zero local dependencies**: Connect directly to `https://axiomlab.online/mcp` with `Authorization: Bearer <API key>`.
* **Plan-before-spend safety**: The assistant uploads your local files (up to 20 MB per file over MCP, or 50 MB via REST), creates a plan (`POST /v1/plans`), shows you the workflow, suggested duration, and format, and **waits for your explicit yes** before starting production (`POST /v1/videos`).
* **QA-checked commercial MP4 output**: Every delivered film includes structured narration, visual chapters or UI/product proof, and frame-accurate captions—and failed or cancelled jobs are never charged.

---

## 1-Minute Setup: Connect Claude Code & Cursor to Axiom Lab MCP

First, create an API key at **[`https://axiomlab.online/dev`](/dev)** under **Developers → Overview → API keys** and export it as `AZIMO_API_KEY`.

### Option A: Claude Code CLI

```bash
claude mcp add --transport http azimo https://axiomlab.online/mcp \
  --header "Authorization: Bearer $AZIMO_API_KEY"
```

### Option B: Cursor (`~/.cursor/mcp.json`)

```json
{
  "mcpServers": {
    "azimo": {
      "url": "https://axiomlab.online/mcp",
      "headers": {
        "Authorization": "Bearer ${env:AZIMO_API_KEY}"
      }
    }
  }
}
```

---

## What You Can Ask Your AI Assistant to Produce

Once connected, point Claude Code or Cursor at files in your repository or desktop and ask in plain language:

| Prompt Example in Claude Code / Cursor | Underlying Axiom Lab Workflow | Deliverable Spec |
| :--- | :--- | :--- |
| *"Turn `./docs/architecture-spec.pdf` and `./screenshots/` into a 60-second project pitch video for enterprise security reviewers."* | **Project Video** ([`/create/project`](/create/project)) | 30–180s · 16:9 1080p MP4 with narrated 4-beat proof & verbatim captions |
| *"Make a 15-second Tech-precision product video for the Pliva Fold Kettle from `./assets/pack-shot.png` and `./spec.md`."* | **Product Video** ([`/create/product`](/create/product)) | 15–30s · 16:9 TVC in `cinematic`, `tech`, or `lifestyle` look |
| *"Build a 9:16 Founder Card in Financial Grid style from `./founder/portrait.jpg`, `./founder/stage.jpg`, and `./founder/facts.json`."* | **Founder Card** ([`/create/name_card`](/create/name_card)) | 30–90s · 9:16 1080×1920 vertical film + 3:4 cover poster |

---

## Built-In Guardrails: Idempotency, Plan Review & LLM-Ready Docs

1. **Mandatory Plan Confirmation**: Starting production, confirming suggested duration after material analysis, and approving budget stops are marked as confirmation-required actions in the MCP schema.
2. **Idempotent Retries**: Repeating a production start call on the same `plan_id` returns the exact same video job rather than double-charging your account.
3. **LLM-Native Documentation**: Point any coding agent at [`https://axiomlab.online/llms.txt`](/llms.txt) or the single-file bundle [`https://axiomlab.online/llms-full.txt`](/llms-full.txt) (or append `.md` to any developer doc URL, such as [`/developers/cli-mcp.md`](/developers/cli-mcp.md) and [`/developers/api.md`](/developers/api.md)) for zero-noise Markdown context.

## Frequently Asked Questions (FAQ)

### What is the remote URL for the Axiom Lab MCP server?

The remote HTTP MCP endpoint is https://axiomlab.online/mcp. Nothing needs to be installed locally—authenticate with your API key via the header Authorization: Bearer <API key> (create a key at https://axiomlab.online/dev).

### How do I add the Axiom Lab MCP server to Claude Code?

Run: claude mcp add --transport http azimo https://axiomlab.online/mcp --header "Authorization: Bearer $AZIMO_API_KEY". Once connected, you can ask Claude Code to turn any local PDF, markdown spec, or folder of product photos into a project, product, or founder-card video.

### Will the MCP server charge my balance without asking me first?

No. Production can only start from a frozen plan (/v1/plans), and the MCP tool definitions instruct the AI assistant to display the plan (workflow, suggested length, format) and wait for your explicit confirmation before starting production (/v1/videos).

### Which video workflows can an AI agent trigger via the Axiom Lab MCP server and REST API?

Agents can list and run all open workflows: Project videos (30–180s 16:9 from PDFs/decks), Product videos (15–30s TVC showcases from product photos), Founder cards (30–90s 9:16 personal-branding films from photos and checkable facts), and Expert interviews.