Atlas Agent Engine

What it does

Atlas Agent Engine is a platform for building, deploying, and governing AI agents in production, built on MongoDB Atlas. It combines agent execution with long-term memory, retrieval, and governance, and agents can reach data held in other systems such as Salesforce, Oracle, Snowflake, S3, and SharePoint.

Teams can adopt the full stack or a single layer, such as memory or governance alone, and can run open-source frameworks like LangGraph. Each tool call runs under a user or agent identity inside an isolated sandbox, and policies set at the org level cannot be overridden downstream. Every model call, tool call, and memory access is traced for audit.

It suits developers and enterprise teams building agents that act on real-time data and need auditability. The product is in public preview and intended for evaluation and prototyping; MongoDB does not recommend production workloads and provides no SLAs or SLOs during preview. Pricing is consumption-based and subject to change.

What it can do

• Long-term agent memory covering semantic, episodic, taxonomic, and procedural memory managed natively.
• Retrieval powered by Voyage AI embedding and reranking models.
• Per-tool-call identity and least-privilege enforcement inside isolated sandboxes.
• Org-level guardrails and policies that cannot be overridden downstream, with human approval routing for risky actions.
• Tracing of every model call, tool call, and memory access for audit.
• Open standards support including MCP, A2A, and OpenTelemetry, neutral across models and frameworks.

Works with

• MongoDB Atlas
• Model Context Protocol (MCP)
• Agent-to-Agent (A2A)
• OpenTelemetry
• LangGraph

Before you choose

• Public preview: MongoDB recommends evaluation and prototyping only, not production, and provides no SLAs or SLOs for availability.
• Pricing is consumption-based runtime, memory storage, and memory retrieval, and is subject to change during preview.
• Users are responsible for agent behavior and compliance with MongoDB's Acceptable Use Policy and Terms of Service, using the Policy Engine for control.

Listed under

• Coding
• MCP and agent tools

Get started

Start with the Get Started tutorial to create and deploy a Hello World agent; sign-up leads to Atlas registration.

1. Create or sign in to a MongoDB Atlas account at the registration page.
2. Follow the Get Started tutorial to create and deploy a Hello World agent on the Atlas Agent Engine.
3. Review the Install & Authenticate guide and the Limitations section before building or migrating an existing agent.

Public preview rates are subject to change. Runtime is billed per 1000 seconds per vCPU, memory storage per 1000 long-term memory docs stored, and memory retrieval per 1000 docs retrieved. Atlas cluster and data service pricing is listed separately on the MongoDB pricing page.

Source: mongodb.com/products/platform/atlas-agent-engine, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• The MongoDB Agent Platform | Atlas Agent Engine | MongoDB ↗www.mongodb.com
• MongoDB Atlas Agent Engine - MongoDB Atlas Agent Engine - MongoDB Docs ↗www.mongodb.com
• Pricing | MongoDB ↗www.mongodb.com

Company website

mongodb.com: product, pricing and policy pages

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.