Which MCP grant lets an agent create an avatar? Write, no paid scope
On Sume's hosted MCP, avatars_create and the avatar video tools are paid tools that need the Write toggle at consent. There is no mcp:paid scope.
To let an agent create an avatar through Sume's hosted MCP, grant mcp:write on the consent page. The avatar creation tool, avatars_create, is listed as a paid tool, and there is no separate mcp:paid scope (as of 2026-10-09). With read-only access, the agent can list and read avatars but cannot spend.
Which avatar tools sit on which side
The MCP tools page splits avatar tools into read and paid. The agent sees read tools with mcp:read, and the paid ones only after you opt in to write.
| Group | Tools |
|---|---|
| Read | avatars_list, avatars_get, avatars_search, avatar-videos_list, avatar-videos_get |
| Paid | avatars_create, avatar-videos_create, avatar-image-to-video_create |
| Paid (preview flow) | avatar-video-previews_create, _get, _regenerate, _generate_video |
| Paid (separate family) | kling-motion-control_create |
OAuth versus an API key
With OAuth, the client sends you to the MCP host consent page, where Read is locked on and Write is off by default. The docs say mcp:read is required and that there is no mcp:paid scope. An API key gives the full hosted tool set, with spend governed by the wallet and admission, as the OAuth page describes. Either way, writes and paid calls must include an idempotency_key, so give each avatar its own key and reuse that key if you retry the call.
What a read-only agent can still do
A read-only agent can list your avatars and read their status, and it can read avatar videos, so it can answer a question like which handles are ready and which videos were made last week. It cannot call avatars_create or the render tools. That split suits a reporting agent with no budget.
If you later enable write, set a budget in your own instructions: no more than one avatar per run and no render above 15 seconds at plus, which is 15 x 0.245 = $3.675. The wallet and admission still decide whether a call can run.
A safe rollout
Start an agent on read only and let it call avatars_list so you can see its view of your handles. Move to write when you have a budget and a prompt that names the handle. The cost for a new avatar is $0.95; a 10-second talking video from it is 10 x $0.184 = $1.84 at standard, so one agent run that creates an avatar and one video is $2.79 at standard.
If your client does not support the OAuth consent page, an API key is the other way to reach the hosted tools. Keep the key out of chat logs and rotate it if it leaks.
Last, remember the cost of a mistake: an avatar created by a misread instruction is $0.95, and a render at max for 30 seconds is 30 x 0.55 = $16.50. Set the instruction so that the agent asks before any max-tier render.
- Ask the agent to report the handle before it renders.
- Prefer previews for videos over 20 seconds.
- Review the job id list at the end of a run.
Sources
Related posts
More in Sume Avatar 1.0
- Moving Company Spokesperson Ad: A 20-Second Avatar for About $4.90
A 20-second plus-quality Avatar 1.0 spokesperson clip costs about $4.90 on Sume, at 98 cents per 4 seconds. English only, with captions at $0.20 extra.
- Introducing Sume Avatar 1.0
Sume Avatar 1.0 is a multi-agent orchestration system as a single avatar model.
- Avatar Face Swap API (Beta): apply an avatar face to a video
Avatar Face Swap 1.0 is a Beta Sume endpoint that applies a ready avatar's face to a short public source video. Required fields, limits, and polling.
- Avatar video previews: approve the first frame before rendering
Create an avatar video preview to get first-frame stills, regenerate them if needed, then call generate-video on the preview id to render the final video.
Written by Sume