Meta Launches Its Enterprise AI Platform and Hires MongoDB's CEO to Run It
Meta launched Meta Enterprise Platform to sell its AI stack to businesses, and hired MongoDB's CEO to run it. MongoDB lost its second CEO in eleven months, one day before its investor day.
What Meta Announced on September 28
The Press Release and Zuckerberg's Thread: "The Next Major Pillar of Our Business"
Meta published the launch of Meta Enterprise Platform on its newsroom, a business unit that brings its full AI stack to enterprises and developers. Mark Zuckerberg signs the text and also introduced it in a four-post thread on X. Bloomberg reports he described the unit as "the next major pillar of our business", while his X post spoke of using "strengths few companies have": advanced models, leading agents, large-scale infrastructure and years working with businesses.
The Four Products on Sale From Day One: Muse Agent, Meta Business Agent, Muse API and Muse Code
The offer opens with four products: the Muse agent, Meta Business Agent, Muse API and Muse Code. The first two target daily operations and customer service; Muse API is the entry point for a team to build on Meta's models, and Muse Code points at people who write software.
What Already Existed Before This Announcement (and What Is Actually New)
It pays not to sell as new what was already there. Muse, Meta Business Agent and the Muse Code beta existed before Monday; what is new is the unit that packages them and puts them on sale together under a single enterprise brand. Meta had already shown the Muse family at its hardware event on September 24, where Meta Connect 2026 unveiled the camera-free glasses and the Muse Charm gadget; that was consumer, this is business.
Muse Code: the Piece Developers Care About
A Terminal and CI Agent: It Plans Tasks, Edits Files and Runs Commands
Muse Code is a coding agent that works in the terminal and in continuous integration. According to its documentation, it can plan tasks, edit files and run commands. It is the format other coding assistants already use, but bundled with the model API in a single package.
User Approvals and an OS Sandbox: What It Promises and What It Doesn't
Two of the most cited pieces are user approvals and an operating-system sandbox. In plain terms: the agent asks permission before sensitive actions and runs in an isolated environment. That is a containment promise, not a guarantee — a sandbox limits the blast radius, it does not remove the risk that a misread instruction does something unwanted inside what is allowed.
The September 22 Beta and the Muse Spark Model Behind It
The Muse Code beta launched on September 22, 2026, with the Muse Spark model behind it. If you plan to try it, treat that detail for what it is: beta software, on a recent model, inside a unit that has just been born.
The Other Side of the Announcement: MongoDB Without a CEO the Day Before Investor Day
Desai Steps Down Effective Immediately After Eleven Months
Chirantan "CJ" Desai, MongoDB's president and CEO, resigned effective immediately the same Monday to join Meta as Chief Enterprise Platform Officer, reporting directly to Zuckerberg. He had been in the role eleven months: he took over in November 2025, succeeding Dev Ittycheria.
Ittycheria Returns as Interim and the Board Starts a Successor Search
MongoDB's board named Dev Ittycheria interim CEO; he led the company as CEO and president from 2014 to 2025, and the board began a formal search for a permanent successor. The company reaffirmed guidance for the current quarter and kept its investor day on Tuesday, September 29, at the Nasdaq MarketSite.
The Market Reaction: -16% Premarket, Double-Digit Session Losses and Guidance Left Untouched
Be precise here, because the published figures do not match. Reports range from -16% premarket at $346.50 after a -2.6% close on Friday, through -15% during the session and "more than 18%" in another source, down to lows of ~26-27% at Monday's worst point. The right way to read it is as a range with a source and a moment of measurement, not a single figure. And do not mix it with the September 2 drop, which was about results: Atlas grew 28.9%, below the ~30% expected, and the stock fell 13.5% that day.
Why Meta Is Doing This Now
A Lot of AI Infrastructure and the Need for a Return That Isn't Advertising
Meta has spent years building AI infrastructure at enormous pace and needs a return path that does not depend on advertising. Selling enterprises the same stack that until now lived inside its products is the most direct way to start billing it. TechCrunch and CNBC frame it that way as third-party analysis, and Gizmodo adds the counterpoint: the launch comes while the company faces criticism over the behavior of its own agent experiences.
Who It Competes With for Enterprise Budget
It competes for the same enterprise AI budget as OpenAI, Microsoft, Google and Salesforce, among others. Meta's difference is that it arrives with its own infrastructure, its own models and a base of businesses already using its commercial tools. The landscape shows up in data like the AI adoption index Anthropic published, which measures how much weight AI carries inside companies themselves.
Hiring a Database CEO Is Not Random: Selling AI to Enterprises Is a Platform and Data Problem
Meta did not go looking for an advertising executive or a consumer product leader: it hired the CEO of a data infrastructure company. That is a signal about where it believes the difficulty lies. Going from home AI to enterprise AI is like going from cooking at home to opening a restaurant: the recipe does not change, everything around it does — licenses, schedules, inspections.
What It Means for Developers
Before You Adopt: SLA, API Pricing, Data Retention, Region and an Exit Clause
Before integrating any of these pieces, review five things: the SLA, usage pricing, where your data is processed and retained, whether there is a regional deployment option, and an exit clause. Meta published no pricing for Muse API or Muse Code in the release, so do not invent it: ask for it.
Working With a Vendor That Just Changed Leadership: What to Review
When a vendor changes leadership, what you review is concrete: the committed roadmap, live contracts, who signs now, and which promises depended on the person who left. Changing a CEO is changing captains mid-flight: the plane keeps flying, but you want to know who has the controls before signing the next contract.
If You Run MongoDB or Atlas: Continuity, Versions, Support and a Contingency Plan
For anyone running a cluster, the message is calm plus preparation: the product and the guidance stay the same for now, because the company reaffirmed them. The sensible move is verified backups, supported versions up to date and a documented migration path. If your database is also slow to respond, the diagnostic work is the familiar one, as in the case of the four-second query in MySQL 8.4.
What Is Still Unknown
There are no statements from Desai or Zuckerberg on the specific reasons for the hire. Nor is there published pricing, quotas or service levels for Muse API or Muse Code, and no general availability dates. On MongoDB's side, who becomes permanent CEO and how long Ittycheria's interim stint lasts remain open.
Conclusion
On Monday the 28th there were two moves that explain each other: Meta turned its AI into a sellable enterprise product and put in charge someone from the database world, while MongoDB lost its CEO a day before its investor day and reaffirmed its numbers. For developers, the useful question is not who wins the headline, but whether the vendor you pick comes with a contract, an SLA, clear retention and a way out. If following these industry decisions with the numbers in order is useful to you, we keep taking them apart day by day on the blog.
