Pull to refresh
Logo
What Changed
Following Why Sign Up
MongoDB launches Atlas Agent Engine for production AI agents

MongoDB launches Atlas Agent Engine for production AI agents

New Capabilities

Unified execution, memory, and governance layer moves agents from demos to deployment

September 29th, 2026: Atlas Agent Engine enters public preview

Overview

Updated Oct 17

MongoDB introduced a platform designed to move AI agents out of the demo phase and into production. Atlas Agent Engine packages persistent memory, retrieval, and governance into one service that runs on top of MongoDB's existing database.

Six systems integrators (Accenture, Capgemini, Cognizant, Infosys, TCS, and Tech Mahindra) are building Agent Engine into their AI offerings, and MongoDB's own sales assistant runs on about 30 agents built on the platform. The service is in public preview on AWS with consumption-based pricing that draws on existing Atlas commitments.

Why it matters

AI agents stall in production without memory and governance. Atlas Agent Engine gives 70,000 enterprises one platform to get them live.

Questions about this story

Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.

No questions yet — be the first to ask.

Key Indicators

70,000+
MongoDB Atlas customers eligible for Agent Engine
Existing Atlas customers can adopt Agent Engine without a new contract.
75%
Share of Fortune 100 companies running MongoDB (MongoDB-reported)
MongoDB says more than three-quarters of the Fortune 100 already uses its platform.
3
Platform products launched at Investor Day
Atlas Agent Engine, MongoDB 9.0, and Atlas Infinite shipped together on September 29.
100+
People added to MongoDB's India R&D team over 18 months
A substantial part of the Atlas Agent Engine team is based in India.
30
AI agents powering MongoDB's internal Holly assistant
MongoDB's sales assistant runs on agents built with Atlas Agent Engine.
95%
Cut in time to research and build outbound prospecting lists
Reps using Holly are about 95% faster at list research.

Voices

Curated perspectives — historical figures and your fellow readers.

Sign up to generate historical perspectives on this story.

Log in Sign Up

People Involved

PS
Pablo Stern-Plaza
Chief Product Officer, AI and Emerging Products at MongoDB
Leading the Atlas Agent Engine launch
 Ashish Kumar
Ashish Kumar
SVP at MongoDB
Leading product and engineering work tied to MongoDB's India R&D organization

Organizations Involved

MongoDB, Inc.
MongoDB, Inc.
Public software company (NASDAQ: MDB)
Atlas Agent Engine is in public preview on AWS; six systems integrators are building it into their AI offerings

MongoDB sells the Atlas managed database platform and is extending it into AI agent infrastructure.
Voyage AI
Voyage AI
AI research company, acquired by MongoDB
Its embedding and reranking models power Atlas Agent Engine retrieval

Voyage AI builds embedding and reranking models that MongoDB says rank among the top performers on the RTEB enterprise retrieval benchmark.

Timeline

February 2025 September 2026

6 events Latest: September 29th, 2026 · 1 month ago
Tap a bar to jump to that date

  1. Atlas Agent Engine enters public preview

    Latest Product Launch

    Unified execution, memory, and governance layer for production agents launches at Investor Day.

  2. MongoDB 9.0 and Atlas Infinite debut

    Product Launch

    Database upgrade and elastic compute-storage option ship alongside the agent platform.

  3. MongoDB raises long-term financial targets

    Corporate Event

    At the same Investor Day, MongoDB raised its long-term financial targets. Management guided for Atlas Agent Engine general availability later in the current fiscal year or early the next.

  4. Six systems integrators join Atlas Agent Engine launch

    Partnership

    Accenture, Capgemini, Cognizant, Infosys, TCS, and Tech Mahindra will build Atlas Agent Engine into the AI offerings they already take to market.

  5. MongoDB reveals internal Holly assistant runs on Agent Engine

    Product Development

    MongoDB's sales assistant Holly runs on about 30 agents built with Atlas Agent Engine, cutting prospecting-list research time by roughly 95%.

  6. MongoDB agrees to acquire Voyage AI

    Acquisition

    MongoDB brings embedding and reranking models in-house to power AI retrieval.

Scenarios

1

Atlas Agent Engine reaches general availability in 2027

Likely Resolves by Q1 2027

Discussed by: Constellation Research, which noted the integration of memory and governance in its launch coverage

Public preview runs through 2026 into early 2027. MongoDB moves Agent Engine to general availability, formalizes pricing tiers, and reports adoption among its existing Atlas customer base. The open-standards design, built on the Model Context Protocol and A2A, lets early customers keep their current models and frameworks. Paysafe, named as an early enterprise building toward production, becomes the first public reference.

2

Preview stalls as teams pick model-specific agent runtimes

Possible Resolves by Q2 2027

Discussed by: Competing cloud and model providers that bundle agent orchestration with their own models

Enterprise teams choose runtimes tied to OpenAI, Microsoft, or Google, the vendors whose models they already use. Atlas Agent Engine stays in preview or sees narrow adoption outside MongoDB's database customer base. MongoDB shifts focus to selling Agent Engine mostly as memory and governance infrastructure rather than a full runtime.

3

Agent Engine wins as memory and governance layer, not runtime

Possible Resolves by End of 2027

Discussed by: MongoDB's own modular positioning, which lets customers adopt memory and governance independently of the runtime

Enterprises keep their existing frameworks but replace stitched-together memory and policy tools with Atlas Agent Engine's native versions. MongoDB captures the memory and governance market even where competitors own the runtime. Revenue grows, but the product's role in the market differs from the launch vision of a full agent runtime.

4

Atlas Agent Engine wins as partner-backed runtime for enterprise AI agents

Likely Resolves by Q2 2027

Discussed by: MongoDB launch partners Accenture, Capgemini, Cognizant, Infosys, TCS, and Tech Mahindra

Six systems integrators are building Atlas Agent Engine into the AI offerings they take to market. Their client delivery muscle and relationships could carry Agent Engine into enterprises beyond MongoDB's existing database customer base, accelerating adoption toward general availability.

Historical Context

3 moments from history that rhyme with this story — and how they unfolded.

2007-2012

Heroku and the platform-as-a-service wave (2007-2012)

Heroku launched a platform that managed application deployment, databases, and add-on services as one system. Developers stopped assembling their own servers, caches, and databases, and instead pushed code to a platform that ran it.

Then

Startups shifted deployment to managed platforms and cut infrastructure teams.

Now

Platform-as-a-service became a standard way to run web applications.

Why this matters now

Atlas Agent Engine makes the same consolidation pitch for AI agents: one governed layer instead of wired-together tools that break when the stack changes.

2014

Serverless computing with AWS Lambda (2014)

Amazon introduced Lambda, letting developers run code without provisioning servers. Scaling, monitoring, and orchestration moved into the platform, so teams no longer managed the infrastructure underneath event-driven applications.

Then

Teams stopped managing infrastructure for stateless workloads.

Now

Serverless became the default for event-driven applications.

Why this matters now

A similar shift is underway for stateful agent workloads, with memory, retrieval, and governance moving into the platform layer rather than custom code.

2010s

The integrated data platform shift (2010s)

Web-scale companies ran separate databases, caches, and search indexes wired together by custom glue code. Document databases like MongoDB absorbed those roles as teams abandoned the stitching work and consolidated on one operational store.

Then

Teams removed custom integration layers from their applications.

Now

Document databases became a default for operational workloads.

Why this matters now

Atlas Agent Engine repeats the playbook: absorb the wiring an agent stack currently needs, and let teams give up the glue code that breaks on every model upgrade.

Sources

(12)
+6
MongoDB Newsroom CRN Constellation Research PR Newswire StockTitan The Cerbat Gem CXO Today Dataquest MongoDB MongoDB MongoDB MongoDB