Runway Brand Kits in MCP, and how Sume handles references
Runway MCP agents can now reference Brand Kits, edited only in Runway's app. Sume's docs list no saved kit; you pass references per request.

Runway's changelog says that on Sep 22, 2026 agents using Runway MCP can reference Brand Kits, while creating and editing a kit stays in Runway's main product. Sume's video docs describe no saved brand kit; you keep your brand look by sending reference images, and for some models video or audio references, with each request.
What did Runway change?
According to Runway's changelog, read 2026-09-29, the Sep 22, 2026 entry "Brand Kits in Runway MCP" lets agents reference Brand Kits, and creation and editing stay in the main product. The entry does not list what a kit contains or which models honor it, so this page does not either.
Does Sume have a brand kit?
The Sume video docs read for this page describe references as fields on each request: frame_images for first and last frames, and input_references for style or content reference images that the model uses as visual guidance. They do not describe a saved kit object. Anything you want repeated across clips, such as a logo lockup, product photo or palette board, goes into every request, or into your own template that builds the request.
If both fields arrive together, frame_images takes precedence and the request is treated as image-to-video. Use input_references alone when you want guidance rather than exact frames.
Which models accept which references?
Only models whose supported_input_references lists a type accept it, so read the catalog before you build a template.
| Model | Reference types the docs state |
|---|---|
seedance-2 | image_url, video_url, audio_url (from the catalog example) |
| Seedance 2.x, Wan 3.0, MiniMax H3, MiniMax H3 Max | Audio and video references are honored |
gemini-omni-flash-1.1 | Up to 10 reference images and up to 3 reference videos of at most 3 seconds each; no reference audio |
How does this work when an agent calls Sume?
Through hosted MCP, generate_video is a paid tool. Under OAuth it needs the opt-in Write scope, and every paid submit needs an idempotency_key. The agent has to be handed the reference URLs, or fetch them from your own store, on every call; nothing in the docs carries a kit from one call to the next.
Practical setup: keep one folder of brand references, note the URL of each, and put the same list in the prompt or the template your agent uses. For a model whose supported_input_references omits a type, drop that type rather than expecting the model to ignore it. The video docs and the Video Router page list the fields.
When would a saved kit be the better fit?
A kit you edit once in a product UI suits a team that wants non-technical people to control the look. Per-request references suit code that already assembles requests, because the brand inputs live beside the prompt in version control. Those are two workflows, and this page does not rank them.
Sources
Related posts
More in Agents
- Run the Sume video agent from your backend with Agent Completions
POST /v1/agent/completions runs the same agent as the Sume Agents chat, with tools and media generation, and returns an async run receipt you poll or webhook.
- Safe automation for AI agents that call paid APIs
Keep agents read-only by default, keep secrets out of logs, and on hosted MCP send an idempotency_key, preview with dry_run, and cap with max_spend_usd.
- Scheduled AI video agent runs: cron, API triggers, and receipts
A Sume schedule is a saved Agents automation that runs on a cron cadence and returns a run receipt. Author it in the dashboard; start and monitor runs by API.
- What is a video agent? How Sume defines and runs one
In Sume's docs, a video agent is a sandbox Agent that composes generation tools into a post-ready video. Brief it in chat, or call it over HTTP.
Written by Sume