Figma’s third-quarter update offers a rare quantified glimpse of how generative-AI features are being used inside larger customer accounts. As of Oct. 31, the company said 65% of paid customers with more than $10,000 in annual recurring revenue were using Figma Agent weekly, while more than 80% were consuming AI credits weekly. The measures are company-reported, not independently audited adoption statistics, and they describe a defined customer cohort rather than Figma’s entire user base.
What Figma reported
In its Nov. 4 release of third-quarter 2026 results, Figma said that, on a weekly basis as of Oct. 31, 65% of its paid customers above $10,000 in annual recurring revenue (ARR) used Figma Agent. More than 80% of that same cohort consumed AI credits weekly. The company also reported that the number of weekly active users consuming credits among these larger paid customers increased almost 50% between July 1 and Oct. 31. These are related but not interchangeable measurements: two are percentages of accounts in a customer segment at a point in time, while the growth statistic counts weekly active users who consumed credits across a four-month interval. Figma’s results announcement is the source for these product-usage figures.
Figma’s release gives two additional measures for its Model Context Protocol (MCP) server. Figma said total MCP usage grew three times quarter over quarter during the three months ended Sept. 30, while write-back usage grew four times quarter over quarter. In October, the company said it improved and expanded the MCP server and launched integrations with 20 partners. The time windows matter: the growth comparison for server use covers Q3 against Q2, whereas the company’s nearly 50% credit-user increase compares July 1 with Oct. 31. The partner integrations are a separate October product update, not a measure of Q3 adoption.
Figma described its MCP server as its fastest-growing product in Q3 when measured by growth in weekly active users. “Fastest-growing” here is the company’s characterization using that growth measure; it does not mean MCP had the largest installed user base or generated the most revenue. The release does not disclose absolute MCP user counts, the number of write-back actions, or the starting point from which either multiple was calculated. A tripling or quadrupling can signal accelerating use, but without those base figures it cannot establish scale.
How to read the customer cohort
The threshold is important. Figma’s release reported 16,566 paid customers with more than $10,000 in ARR as of Sept. 30, up by more than 3,650 from a year earlier. It reported 1,717 paid customers above $100,000 in ARR, more than 450 higher year over year. The AI usage percentages use the larger-than-$10,000 cohort, not only the $100,000-plus cohort and not every person who uses Figma. Figma defines a paid customer as a separately billed customer account with an active paid subscription. One organization can count as multiple paid customers when separately billed divisions, subsidiaries, or teams have their own subscriptions.
That definition means the percentages should be read as account-level enterprise-segment indicators, rather than as the share of individual designers, employees, organizations, or all Figma users. The release does not supply the number of eligible accounts on Oct. 31 alongside the AI percentages, and its count of 16,566 is dated Sept. 30. It is therefore not appropriate to multiply the October percentages by the September account count and present the result as an October customer total. The company’s metric definitions also define “more than $10,000” as $10,000 or more of ARR, a boundary convention worth noting when interpreting the label.
The product terms also describe different behavior. Figma Agent is the company’s AI agent feature, while AI credits are a consumption measure for AI capabilities. The fact that more than 80% of accounts consumed credits weekly and 65% used Agent weekly does not establish that the same users performed both actions, or that Agent alone drove credit use. Figma did not break down weekly consumption by product, customer size, seat count, geography, or use case in this announcement.
MCP and write-back are growing from an undisclosed base
MCP, or Model Context Protocol, is a standard for connecting AI applications to external tools and information sources. Figma’s server makes it possible for supported AI tools to interact with design context and, in some workflows, return changes to Figma. “Write-back” refers to sending changes into Figma rather than only retrieving context. These connections could make design software more useful inside AI-assisted workflows, but the quarterly growth multiples alone do not show how often teams use those workflows or whether the integrations have become routine in production.
The company’s announcement pairs those usage claims with a count of 20 partner integrations launched in October. It does not name the partners in the cited results passage, quantify the contribution of those integrations, or explain how usage is counted across partner tools. A single integration may support several workflows, and an increase in total usage could reflect more accounts, more actions per account, a broader set of integrations, or some combination. Without a disclosed methodology and absolute baseline, readers cannot distinguish among those possibilities.
The announcement also says Figma more than doubled the share of work powered by its own post-trained design model from July 1 through Oct. 31. That is another company-reported measure of deployment mix. The release does not give the share at either endpoint, the amount of work included in the denominator, or a comparison of quality, speed, cost, or customer preference. It supports the narrow conclusion that Figma says its model was used for a larger share of work; it does not on its own prove that the model is better than alternatives or that all AI work is now powered by it.
The adoption figures arrive alongside growth
Figma reported Q3 revenue of $391.4 million, a 43% increase year over year, and said the result was $16.4 million above the high end of its prior guidance. It reported GAAP gross profit of $314.4 million and an 80% GAAP gross margin. Non-GAAP operating income was $47.0 million, while GAAP operating loss was $119.2 million. Those figures frame the adoption claims within a growing software business, but the release does not attribute the quarter’s revenue growth to Agent, credits, MCP, or any specific AI product.
The independent MarketBeat earnings page also lists a Nov. 4 reporting date, Q3 earnings of $0.10 per share, and revenue of $391.42 million. Its revenue figure is close to the company’s rounded $391.4 million figure. MarketBeat verifies the date and financial-results context for this story, but its page does not report the enterprise AI adoption percentages, the credit-user growth, or the MCP and write-back multiples. It is not independent validation of those product metrics.
Figma’s own release said it raised its full-year revenue and non-GAAP operating-income outlook. That is management guidance: a forward-looking expectation rather than a realized result. The company attributed its confidence to customer expansion and investment as well as improvements in AI cost efficiency. Those are management’s explanations, not independent findings that AI adoption caused the quarter’s performance or that cost improvements will continue at the same pace.
What the metrics cannot prove
The numbers establish that Figma says a substantial share of its larger paid accounts were using its Agent and consuming credits weekly at the end of October. They also establish that the company reported rapid growth in MCP and write-back usage against the prior quarter. But they do not tell readers whether the activity was broad across an account or concentrated among a few users, whether use was experimental or tied to completed work, or how much time or money customers saved. The release does not disclose retention, conversion, satisfaction, or productivity results for AI users versus non-users.
There are further limits to comparing these metrics with adoption claims from other software companies. Definitions of “weekly active,” “usage,” “customer,” and “AI credit” can differ. Figma has not published a methodology in the announcement that would let an outside reader reconstruct the percentages from a sample or transaction-level data. Nor does the report establish that the October 31 snapshot will persist into later quarters. The careful description is therefore “Figma reported” or “Figma said,” not that an independent party measured the underlying behavior.
The time span is another reason to avoid treating the release as a clean same-day snapshot of every metric. The headline Agent and credit percentages are explicitly “as of October 31” and expressed on a weekly basis. The near-50% credit-user growth compares July 1 through Oct. 31. The MCP and write-back multiples compare the three months ended Sept. 30 with the prior quarter. The metrics point in the same general direction, but their distinct windows prevent them from being combined into a single adoption rate or trend line.
Why enterprise usage is worth watching
Even with those limits, account-level adoption is more informative than a product announcement that only describes what a tool can do. Figma’s reported weekly measures indicate that AI features are reaching a meaningful portion of its higher-spending paid-account base, rather than remaining solely a demo or a preview. The data are especially relevant because they cover multiple product surfaces: an agent, credit-consuming AI capabilities, and connections through MCP. That breadth may matter to design teams considering whether AI belongs in their day-to-day software stack.
For software vendors, repeated use can matter more than initial sign-ups, because recurring behavior may create opportunities to embed a product in existing workflows. Still, the announcement does not disclose whether Figma charges separately for these features, how credits relate to subscription revenue, or whether use drives upgrades or renewals. Adoption is not the same thing as monetization. Investors and customers should treat the figures as evidence of reported activity, not as proof that those activities are already a significant source of incremental revenue.
The MCP report is also strategically notable because it signals activity beyond a single first-party interface. An MCP server and partner integrations can connect a design product to third-party AI tools. That can make software more interoperable, while also raising practical questions about permissions, data handling, change review, and ownership of generated work. Figma’s earnings release highlights the integration expansion but does not detail those safeguards or offer outcomes from customers deploying the connections at scale. Any conclusion about security or productivity would require further evidence.
What to watch next
Future disclosures could make the present claims easier to interpret by providing absolute weekly active account counts, consistent quarter-to-quarter definitions, and the share of larger accounts that use each feature repeatedly. Usage by itself is an early signal; retention, paid conversion, credit consumption per account, and the economics of serving AI requests would help show whether that signal is durable and commercially meaningful. Figma may or may not publish those details, so they should be treated as questions for subsequent reporting rather than promised disclosures.
For MCP and write-back, useful follow-up evidence would include baseline usage, active-account counts, repeat activity, and clearer definitions of the event being counted. For the post-trained design model, the share of work powered by the model is only one dimension. Quality, customer acceptance, cost, and the extent to which the model is used in production workflows would give the deployment statistic more context. Until then, the reported growth rates support a claim of rising activity, but not a quantified conclusion about impact.
The central takeaway is specific: Figma says that by Oct. 31, weekly Agent use had reached 65% and weekly AI-credit consumption exceeded 80% among its paid accounts with at least $10,000 in ARR; credit-consuming weekly active users in that cohort had risen almost 50% since July 1. It separately says Q3 MCP usage tripled and write-back quadrupled quarter over quarter. Those claims make this a notable enterprise-adoption signal, but the evidence remains company-reported and lacks the absolute counts and independent validation needed to assess depth, causation, or financial contribution.
FAQs
What share of Figma’s larger paid customers used Agent weekly?
Figma said 65% of paid customers with more than $10,000 in ARR used Figma Agent weekly as of Oct. 31, 2026. The measure is account-cohort specific and company-reported.
How quickly did AI-credit activity grow?
The company said weekly active users consuming credits in the above-$10,000 ARR paid-customer cohort increased almost 50% from July 1 to Oct. 31. It also said more than 80% of that cohort consumed credits weekly as of Oct. 31.
What were the MCP and write-back growth figures?
Figma reported that total MCP usage grew three times quarter over quarter during Q3 2026 and write-back usage grew four times quarter over quarter. It did not disclose absolute counts in the results passage.
Did MarketBeat independently verify these AI adoption metrics?
No. MarketBeat’s earnings page corroborates the Nov. 4 report date and reports Q3 EPS and revenue, but it does not independently verify Figma’s AI Agent, credit, or MCP usage claims.
Do these adoption measures show that AI drove Figma’s revenue growth?
No. Figma reported 43% year-over-year revenue growth, but the release does not attribute that growth to Agent, credits, MCP, or another specific AI feature.