figma-to-markdown-mcp
This server acts as a token-efficient bridge between your AI agent and Figma, compacting design context into a minimal plain-text/markdown format.
Fetch compact design context: Call
get_design_context_compactwith a Figma node URL to retrieve a concise, implementation-ready representation including layout, text, assets, and implementation notes — without raw upstream Figma MCP verbosity.Reduce token usage: Internally fetches and prunes full Figma design context, returning only a compact line-based DSL to significantly lower token consumption.
Layout & structure: Frame dimensions, element hierarchy, spacing, padding, border radius, and background colors.
Typography specs: Font family, weight, size, line height, and color for text elements.
Asset references: Image and asset slot references used in the design node.
Traceability: Output includes Figma node IDs, warnings, and fallback hints.
Optional stats: Set
include_stats: trueto append markdown size statistics to the output.Compaction modes & intents: Supports modes (
minimal,balanced,debug) and intent hints (implement,inspect,summarize) to tailor output.Fallback handling: Returns a graceful fallback handoff when compaction cannot be safely completed, allowing the agent to continue with standard Figma MCP tools.
Provides compact plain-text context from Figma design nodes, reducing token usage for LLM agents while preserving implementation-critical facts like layout, typography, assets, and metadata.
figma-compaction-mcp
Current version: 3.0.2
figma-compaction-mcp is an MCP server for Figma-link workflows. It fetches upstream Figma design context internally, prunes it into compact plain-text context, and returns that reduced result to the calling agent instead of the full upstream payload.
What It Is
This project is for teams that want agents to work from Figma node URLs without pushing raw upstream Figma MCP output into the caller model context whenever the bridge can safely handle the request.
The intended flow is simple:
A user gives an agent a Figma node URL.
The agent calls
get_figma_compact_context.This server fetches upstream Figma context internally.
The server compacts the upstream result into a small line-based DSL.
The agent receives compact implementation context and works from that output.
Why Use It
The main reason to use this server is token reduction without losing implementation-critical facts.
Raw Figma MCP responses can be large enough to consume a meaningful part of the caller model context before implementation even begins. This bridge keeps that upstream payload inside the server whenever possible, compacts it first, and only returns the reduced result to the agent.
Lower token usage for Figma-link prompts
Smaller model-context footprint before implementation starts
Cleaner implementation input for agents
Less raw upstream noise in caller context
Traceable output with node ids, typography tokens, asset refs, warnings, and fallback hints
A built-in fallback path when the bridge cannot safely complete
How It Works
This server sits between your agent and the local Figma Desktop MCP server.
User prompt with Figma link
-> Agent calls get_figma_compact_context
-> figma-compaction-mcp connects to local Figma Desktop MCP
-> get_design_context / get_metadata
-> internal compaction
-> compact plain-text context returned to the agentThe public entrypoint is get_figma_compact_context.
figma_url: required full Figma node URLmode: optional compaction mode, one ofminimal,balanced,debugtask: optional intent hint, one ofimplement,inspect,summarizeinclude_assets: optional, defaulttrueinclude_text_specs: optional, defaulttrueinclude_trace_ids: optional, defaulttrueinclude_metadata: optional, defaulttruemax_output_chars: optional explicit output budget
When the bridge succeeds, it returns compact plain-text context plus structured fields for stats, traceability, warnings, and diagnostics. When the bridge cannot safely fetch or compact the node, it returns a fallback handoff so the agent can continue with standard Figma MCP tools directly.
Example compact output:
src|figma|get_design_context|4:5100|FILE_KEY
sum|Example screen|frame|375x876|535,258
el|4:5107|field_card|w343;layout:column;r20;p:16,20,20,20;bg:#ffffff
tx|4:5106|Section title|t1
ty|t1|Inter|600|20|24|#333333
as|imgAsset|asset|4:5107|asset_slot|/assets/example-image.pngExample URL shape:
https://www.figma.com/design/FILE_KEY/FILE_NAME?node-id=NODE_ID&m=dev
Requirements
To use the Figma-link bridge flow, you need:
Figma Desktop
Dev Mode enabled in Figma Desktop
Desktop MCP server enabled in Figma Desktop
Node.js 18+
Default upstream Figma MCP endpoint:
http://127.0.0.1:3845/mcp
Override with:
FIGMA_MCP_URL
Installation
Install globally:
npm install -g figma-compaction-mcpOr run with npx:
npx figma-compaction-mcpMCP Client Registration
Register this server in your MCP client.
Example using npx:
{
"mcpServers": {
"figma-compaction": {
"command": "npx",
"args": ["-y", "figma-compaction-mcp"]
}
}
}Example using a global install:
{
"mcpServers": {
"figma-compaction": {
"command": "figma-compaction-mcp",
"args": []
}
}
}Your client may use JSON, TOML, or another config format, but the command registration model is the same.
How To Use It
Open Figma Desktop and enable Dev Mode and the desktop MCP server.
Register
figma-compaction-mcpin your MCP client.Give your agent a Figma node URL.
Have the agent call
get_figma_compact_contextfirst.Use the returned compact context for implementation, inspection, or summarization.
If the server returns a fallback handoff, continue with the standard Figma MCP tools for the same node.
In practice:
Small and medium components usually return compact context directly.
Large screens can still return larger output when the retained structure, text, and assets matter.
balancedmode is the default for normal implementation work.Only set
max_output_charswhen you intentionally want a hard output budget.
Limitations
Final tool routing still depends on the MCP host or agent. This server can strongly guide usage, but it cannot forcibly override host-side routing.
When the bridge cannot safely complete a request, it returns a compact fallback handoff instead of passing raw upstream payloads through this server response.
Compaction is optimized for implementation relevance, so purely decorative wrappers and chrome-like nodes may be pruned outside inspect-oriented flows.
Other Information
Release history: CHANGELOG.md
Compact contract draft: SPEC.md
Source repository: https://github.com/s9hn/figma-compaction-mcp
Contributions: issues and pull requests are welcome on GitHub
Issues: GitHub Issues
License: MIT
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Latest Blog Posts
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/s9hn/figma-compaction-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server