MCP connector

You need a Studio account with some balance for paid steps — sign up or top up first. Each person connects with their own Studio login, and a connection only ever sees the account that approved it. Connect your own account first; anyone else repeats the same steps with theirs.

ChatGPT

1. 1Open Settings → Apps & Connectors → Advanced settings and turn on Developer mode (needed for custom connectors until the app is listed in the directory).
2. 2Back in Apps & Connectors, choose Create. Name it Stork Content Studio, paste the MCP server URL above, and choose OAuth for authentication.
3. 3ChatGPT opens our sign-in page. Log in (or sign up), choose what the connection may do and its spending limit, then Allow access.
4. 4In a new chat, add the connector from the + menu and ask for what you want. ChatGPT asks you to confirm actions that change something or spend money.

Menu names in ChatGPT change from time to time; look for the place to add a custom connector or app by URL.

Claude

1. 1In Claude, open Settings → Connectors and choose Add custom connector.
2. 2Name it Stork Content Studio, paste the MCP server URL, then Connect and approve on our sign-in page.
3. 3In Claude Code, run the command below, then type /mcp and choose the server to sign in.

Other MCP clients

Any client that supports Streamable HTTP and OAuth 2.1 with PKCE works. It discovers everything from the URL: the server answers an unauthenticated request with a pointer to its OAuth metadata, and clients register themselves automatically (dynamic client registration).

When you approve a connection you decide two things:

• What it may do. View your account, projects, jobs and files (always included); create and edit projects, scripts and imported images; and start paid generation. Leave the last one unticked and the assistant can plan, quote and prepare, but never spend.
• A spending limit for that connection — $5, $10, $20 (the default), $50 or $100 in total, or no connection limit. It sits on top of the budget cap every project has and of your wallet balance; a job that would exceed any of the three is refused before anything is charged.

See and disconnect your connected applications at contentstudio.adventuressmalik.com/oauth/connections. Disconnecting stops the application immediately; reconnecting does not reset what it has already spent.

The connector uses the same pay-per-use prices as the shop, charged to your wallet in US dollars. Nothing is charged until you accept a quote.

• Example: a ten-minute explainer (about 1,500 words) in 20 scenes, each with a generated illustration, costs about $2.60 with a standard voice on the economy tier: 20 × 3¢ of narration plus 20 × 10¢ of illustrations. Assembly is free.
• Quotes are binding. estimate_generation returns an itemised total; start_generation only runs with that quote and the exact total you accepted, and refuses if the script, settings or prices changed since.
• You pay once. Retrying a start with the same idempotency key returns the original job instead of charging again. Anything already generated for the same text, voice and settings is reused at no charge, so editing one scene only pays for that scene.
• Failures are refunded. If a step fails, its price is returned to your wallet automatically; finished steps are kept.

A typical session for one episode of a lecture series, in your own words to the assistant:

“Create a project IME Cosmos — Episode 3: The Expanding Universe in English, widescreen, with a $5 budget. Import these two equation images. Save this script exactly as written: scene 1 … scene 12. Use an illustration for each scene except the equations. Tell me the price before starting.”

1. 1create_project sets the title, language, 16:9 format and the $5 cap.
2. 2import_asset stores each equation as supplied — nothing is redrawn by an image model.
3. 3update_project_script saves the scenes and returns a SHA-256 receipt of each narration, so you can prove the text was kept intact.
4. 4estimate_generation quotes narration, illustrations and assembly, and reports anything that would stop the job.
5. 5After you agree, start_generation returns a job id at once; the work runs in the background.
6. 6get_job_status reports real progress; get_project_outputs returns a download link to the finished MP4.

Every tool declares whether it only reads, whether it can spend, and whether repeating it is safe, so your assistant can ask for confirmation at the right moments.

• Up to 80 scenes per script and 3,000 characters of narration per scene; keep a scene under about 90 seconds of speech.
• 16:9 widescreen (1280×720) or 9:16 vertical. English plus the Indian languages listed by list_capabilities.
• Imported images: PNG, JPEG, WebP or SVG up to 8 MB, uploaded directly or fetched from a public https link. They are shown in full, never cropped.
• Download links are valid for one hour; ask for a new one any time.
• Generated and imported files are kept for 7 days after they were last used or downloaded, then deleted. Download finished videos promptly.

The connector only offers what the studio can really do and price today. These are reported honestly instead of being faked:

Lip-synced presenter video

There is no published per-use price for it yet, so it is not offered. The presenter badge is.

Square (1:1) video

16:9 widescreen and 9:16 vertical only.

Burned-in captions

Narration text is never drawn on screen automatically.

Spoken mathematics

Show equations as imported images and write the narration the way it should be read aloud.

Publishing to YouTube and other platforms

Download the MP4 and upload it yourself.

Cancelling a running job

Unfinished steps of a failed job are refunded automatically.

Your cloned voices and avatars

Use the voices and presenters listed by list_capabilities.

Writing or rewriting scripts

Draft with your assistant; the connector saves and speaks your text verbatim.

• You sign in on our own site; your password never reaches the assistant. The assistant receives a token for this one service, valid for an hour and renewed while the connection is active.
• Every project, file and job is checked against the signed-in account on the server. One connection cannot see another account's work.
• Our AI provider keys stay on our servers. Your script and images are sent to those providers only to produce what you asked for.
• See the privacy policy and terms of service.

MCP endpoint

https://contentstudio.adventuressmalik.com/mcp

Transport

Streamable HTTP, stateless (protocol 2026-07-28; earlier revisions served on the same URL)

Protected resource metadata

https://contentstudio.adventuressmalik.com/.well-known/oauth-protected-resource

Authorization server metadata

https://contentstudio.adventuressmalik.com/.well-known/oauth-authorization-server

OAuth

Authorization code + PKCE (S256), refresh-token rotation, dynamic client registration (RFC 7591), resource indicators (RFC 8707), issuer in the redirect (RFC 9207)

Scopes

studio.read · studio.write · studio.generate

MCP Registry name

io.github.storkmeditator/contentstudio

Server card

https://contentstudio.adventuressmalik.com/.well-known/mcp/server-card.json

Health

https://contentstudio.adventuressmalik.com/mcp/health

The assistant says a voice or stage is unavailable.

That provider is not configured right now. list_capabilities shows what works; pick an available voice. Saving scripts always works.

The estimate says the budget or balance is not enough.

Raise the project's budget, top up your wallet, or generate fewer steps. Nothing was charged.

A download link stopped working.

Links expire after an hour. Ask the assistant for the outputs again to get a fresh link.

The script changed after I asked for a price.

Quotes are tied to the exact script and settings. Ask for a new estimate.

I want to stop an assistant from spending.

Disconnect it at /oauth/connections, or reconnect without the paid-generation permission.

Still stuck? Write to support@adventuressmalik.com.