news / 21 Aug 2026 / 3 min read
Convex Chef is being replaced by its vendor, so it leaves the index
Convex is replacing Chef with a new product and has taken down its entry points. Existing projects still run, so this is not a shutdown, but a builder you cannot sign up for cannot hold a live ranking. Chef is out of the roster, the raw runs, the matrix, both leaderboards it appeared on and both of its compare pages.
By Daniel Osei | Updated 21 Aug 2026
What we found
Convex Chef held rank 17 of 41 in this index with a composite of 70.50. Ahead of this cycle we re-checked its entry points and found the product is on its way out.
https://www.convex.dev/chef now answers with a 307 to https://chef.convex.dev/, which answers with a 302 to https://www.convex.dev/ai?utm_source=chef. That destination page does not mention Chef once. Neither does the Convex pricing page. The docs path docs.convex.dev/chef redirects to a migration guide, and the guide states the position plainly: "Convex Chef has had a good run but we're in the process of replacing it with a new incarnation better suited to long-term application development on Convex."
All of the above was fetched on 20 August 2026.
This is a replacement, not a shutdown
We want to be precise about the difference, because it changes what a reader should do about it. Chef has not been switched off. The same migration guide says "Your Chef projects will continue running but your Chef development flow will be impacted in the meantime", and Convex promises a successor product. If you have a Chef project in production today, it keeps working.
Two things in that guide still matter. The migration is one way: "Chef doesn't have the ability to re-import a project or track any progress from outside it", so exporting a project to local development permanently ends Chef's usefulness for it. And the onward path Convex recommends is not another Convex builder, it is a third party coding agent. The guide names OpenCode and Claude Code.
Why the row had to go rather than carry a note
Our index answers one question: of the products you can start using today, which ones measure best. A ranked row for a builder with no signup entry point does not answer that question, it corrupts it. Rank 17 of 41 is a recommendation whether we label it or not, and we are not willing to make that recommendation for a product whose vendor is walking away from it and pointing users at somebody else's agent.
So Chef is removed from the roster, from the raw runs export, from the comparison matrix, from every leaderboard it appeared on and from both curated compare pages that featured it. Its two builder URLs and both compare URLs now redirect here permanently rather than resolving to a page about a product you cannot buy.
Three things we were publishing that were wrong
The removal also closes three defects in the row itself, and they are worth stating because each one is a mistake another index could make.
The official link on the row pointed at convex.dev, the parent brand. That is a live, healthy, separate product, a reactive backend platform. Publishing the parent URL on a retired sub product silently lends the parent's health to the child's row.
The price we published, 25 EUR per month, was Convex platform pricing wearing Chef's name. Convex Professional is 25 dollars per developer per month plus metered infrastructure, which is a seat and consumption price for a backend, not a builder price for generating an application. Our published wording even dropped the "per developer". Chef never had a price of its own to quote.
And the row carried 45 recorded executions at 60.0 percent strict reliability, the joint worst on the roster. That figure is consistent with a product mid withdrawal, but any execution recorded on or after the redirect went live is not trustworthy, and the run we had scheduled for 21 August 2026 at 09:00 UTC could not have executed against a product with no signup. That scheduled run is cancelled.
What moves as a result
The roster drops from 41 products to 39, because Pazi leaves in the same cycle for a different reason. Every rank below 17 shifts up by one purely because of this removal, not because anything about those products changed. The run count falls from 1,845 recorded prompt executions to 1,755. Every count on the site, both data exports and the sitemaps reflect the roster after the change.
The archived rows are retained by the lab outside the public index. If you need the historical Chef figures for research, write to the lab address on the contact page.