Seedance video agent in ChatGPT and Claude: Sume MCP compared
InfoseekAI put a Seedance video agent into ChatGPT and Claude over MCP on Oct 1, 2026. What a hosted MCP server like Sume's gives you instead, and when.
On October 1, 2026, InfoseekAI announced an AI video agent for Seedance that runs inside ChatGPT and Claude as an MCP-native plugin (openPR release, read 2026-10-05). Its own page describes usage-based, credit pricing with no subscription, and clips of up to 30 seconds on Seedance 2.5 or 15 seconds on other models (InfoseekAI page, read 2026-10-05).
The shape is worth noticing: a video agent is now a remote MCP server that you add to a chat client. Sume's hosted MCP server uses the same pattern, with a different scope. The comparison below is about what each one is, not which is better.
What the Infoseek page says
- Clients: a ChatGPT plugin, Claude through an MCP connector, and custom apps over MCP.
- Inputs: text direction plus image, video or audio references, with resolution, aspect ratio and duration settings.
- Outputs: MP4 files, preview URLs, shareable links and artifact IDs.
- Billing: usage based, depending on resolution and duration.
What Sume's hosted MCP server is
Sume's endpoint is https://mcp.sume.com/mcp. It wraps selected public API capabilities as tools, and the docs tell clients to call tools_list and tools_schema for the live contract rather than assume parity with the HTTP API. Paid tools such as generate_video, tts_create and avatars_create require an idempotency_key, and the router picks a model when you omit payload.model, unless the user named a family.
| Question | Seedance agent plugin | Sume hosted MCP |
|---|---|---|
| Where it connects | ChatGPT plugin, Claude connector, custom MCP apps | Any remote HTTP MCP client, by OAuth or API key |
| Scope | One model family's video agent | Video, image, audio, avatar, crawl, jobs and assets tools |
| Cost control | Credit-based, resolution and duration | Wallet and admission, optional max_spend_usd, dry_run |
| Long jobs | Returned in the conversation | jobs_wait then jobs_result |
When each fits
If you want one model's best-practice prompting inside a chat, a model-specific agent is the shortest path. If the same chat also needs to write a script, voice it, cut a talking-head and check a balance, a general toolset avoids installing one plugin per model. In either case, read the vendor's current page before you rely on a limit: clip lengths and prices in this field changed several times in 2026.
To try Sume's side, follow the MCP quickstart steps and call mcp_health first.
Sources
Related posts
More in Agents
- Sume skipped run with previous_run_active: what it means, what to do
A scheduled Sume run is skipped with skip_reason previous_run_active when the last one is still going. It is not billed work. How to detect it and avoid it.
- Sonnet 5.5 vs GPT-6.1 Sol for video scripts: both $2 in, $10 out
Claude Sonnet 5.5 and GPT-6.1 Sol list the same $2 input and $10 output per million tokens (read 2026-10-05), so pick on fit, not price.
- Spend guard layers for an unattended Sume video agent, in order
Five layers cap a Sume agent that runs unattended: key scope, per-run cap, max_spend_usd and dry_run, generation admission, and the wallet. What each one stops.
- One Sume schedule on a clock and on call: api_trigger_enabled
A Sume cron schedule can also set api_trigger_enabled to accept API runs. trigger_type is fixed at create time, so choose once. Overlap defaults to skip.
Written by Sume