Sora MCP server: why it stopped working and what to use
A Sora MCP server can't make video now: OpenAI shut the Sora API down on September 24, 2026. Point your MCP client at other video models instead.

A Sora MCP server can no longer make videos: OpenAI shut the Sora 2 models and its Videos API down on September 24, 2026, and a Sora MCP server generates by calling that API. To keep generating video from Claude, Cursor or Codex, connect an MCP server that runs other video models, such as fal, Higgsfield, Runway or generate_video on Sume's hosted server. Sume's catalog has no Sora model.
OpenAI's dates come from its help page, deprecations page and API reference; the other servers come from their own pages, and Sume's side from its MCP overview and Video generation docs. All were read on 2026-09-28. For a backend that called the Sora API directly, Sora API alternatives after the shutdown covers the REST move.
When did Sora shut down?
OpenAI's help page says the Sora web and app experiences were discontinued on April 26, 2026, and the Sora API on September 24, 2026. Its API reference for creating a video now says the Sora 2 models and Videos API are no longer available and that no one-to-one replacement API is available. Sora API alternatives after the shutdown has the full deprecation table.
Why does my Sora MCP server fail?
An MCP server that wraps Sora turns the agent's tool call into a create request on OpenAI's Videos API, with a sora-2 or sora-2-pro model. Both models and the API itself are gone, so updating, reinstalling or re-authenticating the server cannot bring generation back. The fix is a server whose tools call a video model that is still served.
If you still have Sora videos to save, the help page points to an export at sora.chatgpt.com/sunset; How to download Sora videos walks through it.
Which MCP servers can replace a Sora MCP server?
Each of these remote servers signs you in to its own account and bills that account. The video models are as each page names them:
| Server | URL | Video models on the page | Account |
|---|---|---|---|
| fal | https://mcp.fal.ai/mcp | A catalog of 1,000+ models, searched with search_models; Kling 3.0 in its pricing example | fal API key or fal sign-in |
| Higgsfield | https://mcp.higgsfield.ai/mcp | Seedance 2.5, Seedance 2.0, Kling 3 and others | Higgsfield account; plan credits |
| Runway | https://mcp.runwayml.com/mcp | Gen-4.5, Seedance 2.5, Kling 3.0, Veo 3.1, based on your Runway plan | Runway account; Runway credits |
| Sume | https://mcp.sume.com/mcp | seedance-2.5, seedance-2-mini, seedance-2, seedance-2-fast, kling-3, wan-3.0, grok-imagine-video-1.5, minimax-h3, minimax-h3-max, gemini-omni-flash-1.1 | OAuth with Write on, or an API key; workspace balance |
How do I replace a Sora MCP server?
Remove the old server from your client, then add one from the table. For Sume in Claude Code, run claude mcp add --transport http sume https://mcp.sume.com/mcp, then claude mcp login sume and turn Write on at consent; paid tools return insufficient_scope on a read-only session. The basics page says hosted MCP still works but is not the primary path today.
Sume's generate_video takes a catalog id from video-router_models, or routes to sume/auto when payload.model is omitted. Old Sora arguments don't carry over as is: seconds becomes duration, size becomes resolution plus aspect_ratio (size returns 400 unsupported_parameter), and input_reference becomes frame_images or input_references with public HTTPS URLs. The migration post maps each field.
What does the replacement cost?
On Sume, video is billed from the workspace USD balance, reserved on submit at provider list × 1.25, plus a 5.5% agent fee by default. Paid calls need an idempotency_key, and dry_run: true previews admission and cost without submitting, so the agent can show the price before it spends. max_spend_usd caps a call when you send it.
Sources
- MCP overview
- MCP quickstart
- MCP tools and gates
- Video generation
- Sume basics
- API reference
- Sume API reference
- OpenAI Help: What to know about the Sora discontinuation (read 2026-09-28)
- OpenAI API: Deprecations (read 2026-09-28)
- OpenAI API reference: Create a video (read 2026-09-28)
- fal: Run MCP (read 2026-09-28)
- Higgsfield MCP (read 2026-09-28)
- Runway MCP (read 2026-09-28)
Related posts
More in Models
- Text to speech in Japanese: set the language to ja
For Japanese text to speech, send the script in kana and kanji and set the language to ja. Left out, Sume can read a kanji-only line as English.
- Text to speech pronunciation: how to fix wrong words
Fix text to speech pronunciation: set the text's language, use a voice recorded in it, and respell tricky words. What Sume's TTS API checks for you.
- Video to anime AI: how to turn a clip into anime
To turn a video into anime with AI, use a video-to-video edit that names the style and what to keep, or restyle one frame and animate it.
- Wan 3.0 vs Seedance 2.5: 30-second clips, inputs and price
Wan 3.0 and Seedance 2.5 both make clips of up to 30 s with audio, frames and references. Wan starts at 2 s and bills per second, Seedance per token.
Written by Sume