Magic Hour MCP Server logo

Magic Hour MCP Server

Create, edit, render, monitor, and retrieve AI-generated videos, images, GIFs, and audio through Magic Hour.

Install Magic Hour MCP Server

Pick your client, then add it in one click or paste the config.

Streamable HTTP (Remote)
Server URL
https://mcp.magichour.ai/

Remote servers are added in the app, not the config file: Customize → Connectors → Add custom connector, then paste this URL.

Hosted endpointLive · 12h agoLive Remote EndpointOur automated health check connected to this remote MCP endpoint successfully.Last checked 12h ago · Nov 16, 2026, 10:15 UTC

Clients with native remote MCP support can connect directly to this URL.

Remote endpoint
https://mcp.magichour.ai/

Overview

Magic Hour MCP Server connects AI clients to Magic Hour’s authenticated media API for generating and editing videos, images, GIFs, and audio. It exposes workflow guidance, upload helpers, asynchronous project polling, account inspection, face detection, and signed-media retrieval alongside creation tools. Use it when an agent needs to turn prompts or uploaded media into finished media projects, then wait for completion and review the returned downloads. Paid generation consumes Magic Hour account credits, while read-only tools support planning, troubleshooting, and project management.

Use cases

  • Generate a video from a text prompt or source image
  • Animate a portrait with an existing audio recording
  • Edit, translate, subtitle, restyle, or upscale a video
  • Upload local media and retrieve completed project files
  • Inspect account credits before starting a paid generation

Key features

  • Asynchronous video, image, and audio project polling
  • Workflow guides with ordered steps and review criteria
  • Presigned uploads for local image, video, and audio files
  • Face detection, selected-face swaps, and character replacement
  • Signed download URLs and inline media retrieval
  • Account, saved-item, and project status inspection

Tools53

What Magic Hour MCP Server lets your agent do. Expand a tool for its parameters.

Verified live~42.6k tokens of context

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked

Recent health check succeeded.

15/15 checks healthy over the last 77d

What it can do

From the 53 tools this server returned when we ran it.

Can delete or modify data · 5Writes files · 1Handles credentials · 1Makes network requests · 44Read-only tools · 47
Tool definitions changed on 2026-11-16. Review them if you approved this server earlier.

Observed 2026-11-16 in an isolated sandbox. A description of reach, not a security audit.

Supply-chain signal

2 high-severity advisories on record for this package. Most advisories affect transitive dependencies and may not be exploitable in this server's actual usage — this is a directional signal, not a security audit.

Critical 1High 1Medium 0Low 0

Scanned 9/1/2026 via OSV.dev · https://api.magichour.ai (npm)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars10Stargazers on the source repository.
Last commitTodayMost recent push to the default branch.
Availability100%Our rolling endpoint + install checks that succeeded.
Tools exposed53Callable tools this server registers over MCP.
Directory activity7 viewsConfig copies, upvotes, and views on AllMCPs.

How Magic Hour MCP Server works

What Magic Hour MCP Server does

Magic Hour MCP Server gives MCP-compatible agents access to Magic Hour media workflows. Its tools cover video generation and editing, image generation, audio projects, uploads, face detection, project status, account details, and completed-file retrieval. Creation operations return a project ID and credit information while processing continues in the background.

Supported workflows include text-to-video, image-to-video, animation, talking photos, spoken advertisements, lip synchronization, video translation, video restyling, subtitle creation, upscaling, character replacement, and face swaps. Image and audio projects can also be created and monitored through the available project tools. Face-related operations are restricted to a user’s own likeness or content they are authorized to use, and must not be used for impersonation, deception, sexual content, or content involving minors.

How it works

The server builds MCP tools from an OpenAPI specification when it starts. Agents can call get_workflow_guide for ordered inputs, creation-tool names, example requests, follow-up steps, and review criteria for supported workflows. The guide helps plan a job but does not test model access, account eligibility, or the likely cost of a particular request.

Media generation is asynchronous. A creation tool returns an ID immediately; the agent then calls the relevant wait_for_*_project tool or status tool. A timeout does not mean the project failed. The same project ID should be used to resume polling rather than starting another generation. Completed jobs provide signed download URLs, which should be used exactly as returned.

Local files use a separate upload sequence. Call get_upload_urls, send the file bytes to the returned presigned URL, and pass the resulting file_path into a creation tool. The hosted server does not read local filesystem paths supplied by a caller. Inline image, audio, and video retrieval is limited to 8 MiB, so larger outputs should be downloaded from their signed URLs.

Setup and configuration

The hosted creation endpoint is https://mcp.magichour.ai/. A local development instance can be installed with pip install -e . and started with python main.py; its MCP endpoint is http://127.0.0.1:8000/. Local development serves the root path rather than /mcp.

By default, API requests target https://api.magichour.ai. MAGIC_HOUR_API_BASE_URL can point to a mock or alternate API base, and MAGIC_HOUR_OPENAPI_PATH selects the OpenAPI specification file. Production OAuth configuration uses MCP_OAUTH_ISSUER_URL and MCP_OAUTH_RESOURCE_URL, both set to the hosted MCP service URLs in the standard deployment.

Hosted Claude and ChatGPT connectors handle authorization after sign-in and consent. Manual bearer clients should send a Magic Hour API key in the Authorization header. Calling get_account_info is the recommended way to verify API access before starting a paid project; ping only checks server reachability.

Tools and capabilities

  • Plan supported workflows with get_workflow_guide and inspect the public workflow resources.
  • Create and edit media with video, image, audio, animation, talking-photo, ad, translation, subtitle, restyling, upscaling, character, and face tools.
  • Upload source media through presigned URLs instead of placing raw bytes in MCP arguments.
  • Poll video, image, and audio projects until completion, failure, cancellation, or timeout.
  • Inspect account credits, subscriptions, saved characters, references, voice recordings, moodboards, and brand kits.
  • Retrieve completed media inline for compatible clients or through the exact signed download URLs.

Generation uses Magic Hour credits. Technical failures generally return credits after processing, but unsatisfactory completed results do not automatically receive a credit return. A workflow’s guide and account information are not a request-specific price or eligibility quote, so model, source-media, and plan access should be checked before spending credits.

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about Magic Hour MCP Server

Magic Hour MCP Server is an MCP server that connects AI clients to Magic Hour’s media API. Its main tools create and edit videos, images, GIFs, and audio, upload source files, poll asynchronous projects, inspect account details, detect faces, and retrieve completed media through signed download URLs.

Install the local package with `pip install -e .`, then start it with `python main.py`. The local MCP endpoint is `http://127.0.0.1:8000/`. You can also connect an MCP client to the hosted endpoint at `https://mcp.magichour.ai/`.

Call `get_upload_urls`, upload the raw bytes to the returned presigned `upload_url`, and pass the returned `file_path` to the creation tool. Raw file bytes and local filesystem paths are not placed directly in MCP tool arguments.

Yes. Claude Desktop is a general-purpose MCP client that can connect to the server. The hosted endpoint is `https://mcp.magichour.ai/`; authorization is handled through the hosted connector’s sign-in and consent flow.

Call `wait_for_video_project` with the project ID returned by a video creation tool. If polling times out, call it again with the same ID instead of creating another project; completed jobs return signed download URLs.

Related MCP servers

Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewMagic Hour MCP Server AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/magic-hour-mcp-server?style=directory)](https://allmcps.com/mcp/magic-hour-mcp-server)
HTML Embed
<a href="https://allmcps.com/mcp/magic-hour-mcp-server"><img src="https://allmcps.com/api/badge/magic-hour-mcp-server?style=directory" alt="Magic Hour MCP Server on AllMCPs" /></a>