Figma announces new features to define how agents work on its platform
What did Figma announce for AI agents, and what is Figma Specs? On 14 October 2026, at its first Config event in India, Figma announced Figma Specs: a Markdown-based editor where teams write down thei
What did Figma announce for AI agents, and what is Figma Specs?
On 14 October 2026, at its first Config event in India, Figma announced Figma Specs: a Markdown-based editor where teams write down their design language so that both people and AI agents can follow it. It can be a standalone file or sit inside the canvas.
Specs can link canvas elements, FigJam diagrams, code layers and designs. Teams can also add design libraries, code libraries and guidelines into one shared design-system context that agents check for brand rules. Figma's reasoning is that AI output is non-deterministic, so it needs steering towards more predictable results.
What else changed around agents and code?
Code layers, which let an agent turn a design into code or edit part of the code to iterate on a prototype, moved from private beta (announced in June 2026) to open beta. Figma also added more than 20 new tools to its MCP server, including Open Code, Dia, Meta Muse and Sarvam.
This builds on 24 March 2026, when Figma opened the canvas to agents. The MCP server's use_figma tool lets agents create and edit Figma files and components linked to a team's design system. Supported clients include Claude Code, Codex, Cursor and Copilot CLI.
Why does this matter if I run a small business?
It matters because the bottleneck is moving. Figma's chief design officer Loredana Crisan said 17 times more lines of code are being produced than in the past, while releases have grown only 1.3 times. Making things is getting cheap. Making them consistent and shippable is not.
For a founder, that means an agent can now draft pages quickly, but without written rules it will draft them differently each time. Your brand becomes something you must document, not something that lives in one designer's head.
Do I need Figma Specs to get consistent AI-generated design?
No. The useful idea is not the product but the habit: write your brand rules in plain text that a machine can read. Figma's skills are simply Markdown instruction files that anyone can write without a plugin or code. You can adopt the same approach in any tool.
A one-page brand brief is enough to start. Include:
- Colours with exact hex values and where each is used
- Fonts, sizes and spacing rules
- Button, form and menu patterns
- Tone of voice, with three examples of good copy and three of bad
- What the agent must never change, such as prices or opening hours
What will it cost, and should I budget for it?
Agent-on-canvas is free during the beta, but Figma says it will become usage-based and paid, and describes the MCP server as a future paid API. Treat today's free access as a trial, not a pricing model.
Do not build a workflow that only works if it stays free. Keep your brand rules in a file you own, so you can move between tools if the price changes.
What is Braynex Services' view?
Our view at Braynex Services: the real risk is not falling behind on AI design tools, it is building your business inside someone else's platform. Specs written in a vendor's editor are convenient, but they are rented, like any other platform.
We see the same pattern in small businesses. A nail salon we worked with paid Fresha about 1,800 a month in commission, then moved to its own booking system at 35 a month, saving roughly 21,000 a year. Ownership of the customer, data and terms is what compounds.
So keep your brand brief as a plain Markdown file, store it with your website, and let any agent read it. Your website should be the product of your rules, not a by-product of one vendor's roadmap.
What should I do this week?
- Write your one-page brand brief in Markdown and store it somewhere you control.
- Test an agent on one page, such as a menu or services page, and compare the output against your rules.
- Note which corrections you repeat, then add them to the brief.
- Set a monthly cost ceiling for AI tools before beta pricing arrives.
- Check that your site is a real, owned asset first. A live web menu lifted one Manchester restaurant's online orders by 34% within 4 weeks, and no agent can help a business that has no site to improve.
If you would like help turning your brand rules, website and bookings into infrastructure you own, book a free audit at braynexservices.com. Braynex Services will show you where you are renting, where you are losing money, and what to fix first.
Sources
Want this built for your business?
We build the digital infrastructure behind everything you read here. Book a short scoping call and we'll show you what to fix first.
Book a scoping call →