Mediawork

Mediawork MCP server

Read-only Model Context Protocol access to everything Mediawork publishes publicly: the directory of post-production and distribution vendors, the FAQ, the blog, and subscription plans.

Connector URL: https://mcp.mediawork.io/mcp — Streamable HTTP, no authentication, no sign-up.

Listed in the official MCP Registry as io.mediawork/mediawork.

Connect it

Claude Code:

claude mcp add --transport http mediawork https://mcp.mediawork.io/mcp

Anything that takes a JSON config:

{
    "mcpServers": {
        "mediawork": {
            "type": "http",
            "url": "https://mcp.mediawork.io/mcp"
        }
    }
}

Tools

Tool	What it does
search	One query across the directory, FAQ and blog, returning {id, title, url}
fetch	Read one record in full, by an id from search
search_facilities	The vendor directory, filtered by service and place
get_facility	A vendor's full public profile
list_services	The service taxonomy, for resolving a filter
list_places	Cities, states and countries with vendors in them
search_faq	Published FAQ entries
list_blog_posts / get_blog_post	The blog
get_plans	Subscription plans and pricing

search + fetch are the generic pair some clients require; the dedicated tools give richer, filterable results where a client can use them.

Directory questions usually chain: list_services or list_places to resolve a filter, then search_facilities, then get_facility. Every result carries its canonical www.mediawork.io URL so an agent can cite the source.

Read-only, structurally

The deployment holds exactly one environment variable, APP_API_URL, and every byte it serves comes from https://app.mediawork.io/api/www/* — the same public contract the marketing site consumes. There are no database credentials and no write path, so "read-only" is a property of the deployment's shape rather than a promise in a comment.

Contributing

Setup, the test commands and the house conventions are in CONTRIBUTING.md.

Licence

MIT © Screen Arts Limited.