Editor MCP gives AI tools write access to your design system — documentation, tokens, components, contexts, skills, and more.
Who is it for
Editor MCP is for people who already maintain the design system and want to connect their AI agent to Supernova for assistance.
Viewers cannot connect to Editor MCP. During sign-in they can only pick a context, which connects them to Consumer MCP.
You get Editor MCP when you sign in and select a design system. Selecting a context always gives you Consumer MCP, including for Admins (e.g. when they want to test what MCP returns to their design system consumers). The two cannot be combined in a single connection.
How to connect
You need a Supernova account and at least Contributor access or above. Add Supernova as a custom MCP server, then pick a design system when you sign in.
https://mcp.supernova.io/mcp
- Open Settings / Customize
- Navigate to MCPs
- Select New MCP Server
- Paste:
{
"mcpServers": {
"supernova": {
"url": "https://mcp.supernova.io/mcp"
}
}
}
Sign in as Contributor or above, and pick a design system.
Claude Code
Add the server with the Claude Code CLI:
claude mcp add --transport http supernova https://mcp.supernova.io/mcp
On first use, Claude Code opens a browser window so you can sign in. Pick a design system. After authenticating, Claude Code reuses the session for later calls.
Claude for Desktop / Claude.ai
- Navigate to Settings / Customize / Connectors
- Click on Add → Add custom connector
- Integration name: Supernova
- Integration URL: https://mcp.supernova.io/mcp
- Sign in as Contributor or above, and pick a design system
- Use shortcut CTRL/CMD + SHIFT + P and select MCP: Add Server
- Select HTTP (HTTP or Server-Sent Events)
- Enter https://mcp.supernova.io/mcp
- Name it Supernova and start the server
- Sign in as Contributor or above, and pick a design system
codex mcp add supernova --url 'https://mcp.supernova.io/mcp'
Sign in through the browser on first use, and pick a design system.
- To add custom connector, you need to enable Developer mode first
- Go to Settings / Security and login / Developer mode
- Go to Plugins, and click on + in top right corner
- Integration name: Supernova
- Integration URL: https://mcp.supernova.io/mcp
- Sign in as Contributor or above, and pick a design system
To confirm you have Editor MCP, ask the assistant to list tools. You should see sn_admin_write.
If you only see read tools, most likely you picked a context — or you are signed in as a Viewer.
What Editor MCP can do
Editor connections include every tool from Consumer MCP, so the assistant can still search pages, look up tokens, and read components. On top of that, it can make edits in the design system.
Writes go through one tool, sn_admin_write, with one action per call. The assistant must read current IDs and content first. A bundled use-supernova skill on this connection explains how to do that safely.
The consent screen grants read access to your profile, workspace, and selected design system, and write access to design system data.
|
Product area |
What the assistant can do |
|---|---|
|
Documentation |
Create, rename, move, and delete pages and groups. Write Markdown and Supernova blocks, upload images, link Figma frames, set approval, and publish on request. Publishing, including Preview, requires explicit confirmation. |
|
Design tokens |
Create, update, and delete tokens, token groups, themes, and token properties. Includes bulk edits, theme overrides, and property updates. |
|
Components and Figma data |
Create, update, and delete components and their properties. Link a Figma component or Figma data source, and reimport a connected source. Cannot create anything that originates outside Supernova, including Figma components, code components, and Storybook stories. |
|
Contexts |
Create, update, and archive contexts. |
|
Skills and files |
Create, update, and delete skills and knowledge files. |
|
Code automation |
Create, update, delete, and run export pipelines. |
|
Feedback |
List, resolve, and delete feedback. |
What it cannot do
Editor MCP cannot:
- Run when you pick a context at sign-in — that connection is Consumer MCP
- Bypass the signed-in user’s permissions
- Create, delete, or administer workspaces, design systems, or versions
- Manage users - invite, role update, remove from workspace
- Create or directly edit assets, Figma components, code components, or Storybook stories
- Configure authentication for Figma or other external services — connect those in Supernova first
Destructive documentation deletes and publishing need an explicit confirmation in the conversation.
Sample prompts
|
Prompt |
What it does |
|---|---|
|
"Document the Button component from its Figma component set, including anatomy, variants, and states." |
Reads the Figma set and writes a component page covering anatomy, variants, and states. |
|
"Write usage guidelines and do/don't pairs for Dialog, matching the tone of other component pages." |
Reads existing component docs and adds usage guidance in the same voice. |
|
"Replace every mention of Legacy Blue with Core Blue in body copy." |
Finds matching pages and rewrites copy. |
|
"Reorganise our documentation so every component page follows Overview, Anatomy, Usage, Props, Accessibility." |
Checks component pages and adds or reorders those sections where they're missing. |
|
"Mark Tag as deprecated, add a notice to its page, and link Chip as the replacement." |
Updates the component and its documentation page. |
|
"Set Token set to Semantic for every token in the Foreground group." |
Bulk-updates that property on the tokens in the group. |
|
"Set the maturity property on every component: stable, beta, or deprecated." |
Reads each component and sets the property. |
|
"Create a context for the iOS team that includes Foundations documentation and color tokens." |
Creates a context and sets what it includes. |
Available tools
These tools are only on Editor MCP. The assistant picks the tool — you don't need to name it. Writes go through sn_admin_write, one action per call.
Write
|
Tool |
Description |
|---|---|
|
sn_admin_write |
Create, update, and delete documentation, tokens, components, contexts, skills, files, and pipelines. Resolve or delete feedback. One action per call. |
Scope helpers
|
Tool |
Description |
|---|---|
|
sn_admin_get_current_context |
Current workspace, design system, writable version, and dataset. |
|
sn_admin_get_current_version |
Current writable version of the design system. |
|
sn_admin_get_current_brand |
Current dataset in this session. |
|
sn_admin_get_brand_list |
List datasets in the design system. |
|
sn_admin_get_entity_url |
Build a Supernova app URL for a known ID. |
|
sn_admin_get_documentation_block_schema |
Authoring schema for documentation blocks before writing them. |
|
sn_admin_get_documentation_publish_history |
Recent documentation publish builds. |
|
sn_admin_get_export_pipelines |
Export pipelines in the design system. |
|
sn_admin_get_export_pipeline_runs |
Recent runs for a pipeline or exporter. |
|
sn_admin_get_export_build_logs |
Build logs for an export run. |
Editor MCP also has access to every tool listed under Available tools on Consumer MCP.
Limitations
Uploading attachments
Whether attachments can be uploaded depends on where your agent runs, not on which tool you use.
Local agents (Claude Code, Codex CLI, Cursor or an IDE agent running on your machine) upload files directly to Supernova storage. All file types and sizes work, up to your plan limit.
Cloud agents (web-based or sandboxed environments, like ChatGPT or Claude.ai) usually cannot make outbound upload requests. Files must be passed inline instead, which caps them at 64 KB. Anything larger fails. If you hit this limit, either re-run the task from a local agent or upload the file in the Supernova web app first, then reference it from the agent.