ChatGPT Desktop App Adds WebMCP Support; OpenAI Introduces Premium Seat for Business Users — BigGo Finance

OpenAI announced on August 25 that the built-in browser in its desktop ChatGPT app now supports “WebMCP,” a standard designed to deepen integration between websites and AI. This makes it easier for ChatGPT and the coding assistant tool Codex to automatically execute tasks such as product purchases, seat reservations, and information searches on WebMCP-enabled websites. The same day, the company also added a premium seat tier to its enterprise plan “ChatGPT Business” and revealed plans to reinstate the five-hour usage limit on the $20-per-month (approximately 3,200 yen) “ChatGPT Plus” plan, which had been temporarily suspended.

Strengthening Web Integration with WebMCP

WebMCP is a standard that communicates to AI how to use functions embedded in websites, such as “purchasing products,” “reserving seats,” and “searching for information.” With the desktop ChatGPT app’s built-in browser now supporting WebMCP, when users visit a compatible site, ChatGPT or Codex can automatically leverage that site’s functions to complete tasks.

Additionally, “ChatGPT Sites,” a feature that allows users to build websites within ChatGPT, now supports creating WebMCP-compatible sites. Users can simply instruct ChatGPT to “build a WebMCP-enabled site with such-and-such functionality” and the site can be created and published. OpenAI has published examples of WebMCP-compatible sites on its developer portal.

The company is also hosting a contest for WebMCP-compatible sites called the “WebMCP Challenge.” The deadline is 5:00 a.m. Japan time on September 4, 2026. The top 10 entries will receive a $3,000 prize (approximately 477,000 yen), a one-year ChatGPT Pro subscription, one “Codex Micro” AI device, OpenAI original merchandise, plus additional prizes from Shopify, Google Chrome, Netlify, Cloudflare, Vercel, and Render.

Premium Seat and Management Features for Business

OpenAI has added a “Premium Seat” tier for high-usage members to its enterprise paid plan “ChatGPT Business.” The Japan pricing page lists it at 19,250 yen per month, with discounted pricing available for annual contracts. It is not a separate standalone plan but rather a seat type assigned per user within a Business workspace.

Premium Seat offers five times the usage allowance of a standard seat and removes the five-hour usage cap. Usage allowances reset on a weekly basis, and if the limit is reached, the workspace owner can add shared credits. OpenAI envisions Premium Seat for workloads that tend to consume large amounts of AI resources, such as analyzing sales reports and customer feedback, building marketing campaigns, and development with Codex on large codebases.

Standard seats and Premium Seats can coexist within the same workspace. Workspace owners can upgrade seats or reassign them to other members based on job responsibilities and usage volume. There is no need to create a separate contract or workspace for Premium; billing, usage status, and spending limits can all be managed collectively. ChatGPT Business requires a minimum of two paid seats, and a combination of one standard seat and one Premium Seat satisfies this requirement. This enables even small startups to configure Premium access only for members with high AI usage, such as developers or data analysts.

Furthermore, OpenAI announced an “Admin plugin” for ChatGPT Work and Codex. Workspace administrators can complete routine tasks directly from the ChatGPT Work or Codex conversation interface, including adding or removing members and groups, reviewing and changing permissions, analyzing usage and credits, and generating reports. The plugin also supports visualizing usage trends in charts, approving or denying access requests, and creating presentation materials to share on Slack.

Five-Hour Limit Returns for Plus

Meanwhile, the five-hour usage limit was reinstated on August 25 for Codex and ChatGPT Work users on the $20-per-month ChatGPT Plus plan. Thibault Sottiaux, who oversees Codex and ChatGPT engineering at OpenAI, confirmed the change on X (formerly Twitter).

For approximately six weeks starting July 13, the company had temporarily suspended the five-hour limit in conjunction with the rollout of its high-efficiency model “GPT-5.6 Sol,” allowing usage based solely on weekly allocations. During this period, OpenAI implemented early resets of weekly usage allowances on multiple occasions to coincide with milestones such as gaining one million additional users, but that grace period has now officially ended.

Sottiaux cited two reasons for reinstating the limit: dividing usage into five-hour blocks helps distribute computational load and prevents server strain from large-scale simultaneous access, and it prevents new or infrequent users from exhausting their entire weekly allocation in a single long session.

The five-hour limit does not mean five hours of actual usage time; rather, it refers to the amount of messages and computing usage provided in five-hour increments. The rate of consumption varies depending on the model, task complexity, input context, and reasoning level. ChatGPT Work and Codex share usage and credit limits, and users who exhaust their base allocation can purchase additional credits to continue working.

The restriction applies preferentially to the Plus plan. The $100-per-month (approximately 16,000 yen) Pro 5x plan and the $200-per-month (approximately 32,000 yen) Pro 20x plan will not be subject to the five-hour limit for the next several months. Enterprise and Edu accounts are also unaffected, with their existing credit-based systems remaining in place.

In conjunction with reinstating the limit, OpenAI reset the current weekly usage for all affected accounts. The next weekly reset is scheduled for September 1.

Users See It as a Push Toward Pricier Plans

According to reports from South Korean media, the reinstatement of the five-hour limit—while the higher-priced Pro plans remain exempt—has drawn backlash from users who question whether the move is “a tactic to push subscriptions to higher-tier plans or additional credit purchases.”

OpenAI introduced “banked resets” in June, allowing users to restore their weekly limits, and this month some Plus users who exhausted their weekly usage saw an $8 (approximately 1,300 yen) reset button appear. However, no evidence has been confirmed that the company intentionally reduced allowances or increased consumption rates.

Sottiaux explained that “the five-hour limit allows us to distribute computational load while maintaining sufficient weekly usage allowances.”