MongoDB Atlas

Parent: MongoDB Expert Knowledge · researched 2026-05-25· 0 sources · 0 concepts · skill mongodb-atlas-expert

Generated from docs/mongodb-atlas-expert-context.md in 10gen/mdb-tam. Use it as a MongoDB Atlas platform reference when planning Atlas architecture, automating Atlas administration, connecting applica

MongoDB Atlas Expert

Generated from docs/mongodb-atlas-expert-context.md in 10gen/mdb-tam. Use it as a MongoDB Atlas platform reference when planning Atlas architecture, automating Atlas administration, connecting applications, designing Atlas-backed schemas and queries, or reviewing Atlas operational posture. Start from the context below, then defer to the linked official MongoDB Atlas docs, MongoDB Manual, and driver docs as the source of truth for exact endpoint, command, operator, and version details. For a deep sub-area, match the task to the Sub-skill routing table below and read the listed references/…md file before answering. [source]

Sub-skill routing table

This skill consolidates 27 Atlas sub-skills as on-demand references - match the task to the table and Read the listed references/…md file before answering deep questions. Do not rely on this table alone for depth. [source]

Atlas platform and architecture

Atlas docs home: deployment types, regions, access control, connection, [source]
alerts, optimization entry points [source]
<https://www.mongodb.com/docs/atlas/> [source]
Atlas Architecture Center: Atlas well-architected guidance across [source]
operational efficiency, security, reliability, performance, and cost [source]
<https://www.mongodb.com/docs/atlas/architecture/current/> [source]

Atlas administration and automation

Atlas Admin API v2 reference: canonical Atlas administration endpoint [source]
inventory and auth model [source]
<https://www.mongodb.com/docs/api/doc/atlas-admin-api-v2/> [source]
Configure Atlas API access: service-account/API-key setup, IP access list [source]
behavior, and the REST/data-plane boundary [source]
<https://www.mongodb.com/docs/atlas/configure-api-access/> [source]
Atlas CLI docs: command-line Atlas management surface and setup flow [source]
<https://www.mongodb.com/docs/atlas/cli/current/> [source]
Atlas Terraform provider guide: infrastructure-as-code surface for Atlas [source]
provisioning and lifecycle management [source]
<https://www.mongodb.com/docs/atlas/terraform/> [source]
Atlas Kubernetes Operator: Kubernetes control-plane integration and [source]
<https://www.mongodb.com/docs/atlas/operator/stable/> [source]

Connection, security, and access control

Connect to a database deployment: connection prerequisites, private [source]
networking choices, firewall requirements, and data-plane connection flow [source]
<https://www.mongodb.com/docs/atlas/connect-to-database-deployment/> [source]
Atlas IP access list: project-scoped client network allow-list behavior, [source]
limits, temporary entries, and CLI entry points [source]
<https://www.mongodb.com/docs/atlas/security/ip-access-list/> [source]
Atlas database users: Atlas-vs-database-user boundary, auth mechanisms, [source]
role model, and operational limits [source]
<https://www.mongodb.com/docs/atlas/security-add-mongodb-users/> [source]

Data access and MongoDB application design

MongoDB drivers: official application client surfaces by language [source]
<https://www.mongodb.com/docs/drivers/> [source]
MongoDB data modeling: access-pattern-first schema design, embedding vs [source]
referencing, and flexible-schema guidance [source]
<https://www.mongodb.com/docs/manual/data-modeling/> [source]
MongoDB indexes: index types, write/read tradeoffs, Atlas UI/CLI index [source]
management, and Performance Advisor entry points [source]
<https://www.mongodb.com/docs/manual/indexes/> [source]
MongoDB aggregation: preferred aggregation-pipeline model and stage-based [source]
<https://www.mongodb.com/docs/manual/aggregation/> [source]

Atlas search and AI surfaces

Atlas Search: full-text search, analyzers, mappings, $search, [source]
$searchMeta, pagination, faceting, and autocomplete [source]
<https://www.mongodb.com/docs/atlas/atlas-search/> [source]
Atlas Vector Search: ANN/ENN vector search, hybrid search, RAG, automated [source]
embedding, and version availability [source]
<https://www.mongodb.com/docs/atlas/atlas-vector-search/> [source]

Streaming, encryption, and newer surfaces

Atlas Stream Processing: continuous stream processing using [source]
aggregation-pipeline syntax over Atlas and Kafka sources [source]
<https://www.mongodb.com/docs/atlas/atlas-stream-processing/> [source]
Queryable Encryption: query encrypted fields without exposing plaintext; [source]
equality/range GA, prefix/suffix/substring in preview on 8.2+ [source]
<https://www.mongodb.com/docs/manual/core/queryable-encryption/> [source]
Atlas AI Integrations: consolidated RAG, agent memory, and embedding [source]
pattern documentation [source]
<https://www.mongodb.com/docs/atlas/ai-integrations/> [source]

Operations, backup, observability, and limits

Atlas alerts: alert conditions, lifecycle, acknowledgement, and org/project [source]
<https://www.mongodb.com/docs/atlas/alerts/> [source]
Atlas cluster metrics: health metrics, real-time metrics, search metrics, [source]
and operator-facing signals [source]
<https://www.mongodb.com/docs/atlas/monitor-cluster-metrics/> [source]
Atlas Cloud Backup overview: backup enablement, redundancy, compliance, [source]
restore-role requirements, and topology caveats [source]
<https://www.mongodb.com/docs/atlas/backup/cloud-backup/overview/> [source]
Atlas limits: component, connection, and topology limits that shape [source]
designs and operating posture [source]
<https://www.mongodb.com/docs/atlas/reference/atlas-limits/> [source]
Atlas Performance Advisor: slow-query analysis, index suggestions, and [source]
read-vs-write tradeoff reminders [source]
<https://www.mongodb.com/docs/atlas/performance-advisor/> [source]

Atlas quick rules

Treat Atlas as the control plane and MongoDB drivers / mongosh as [source]
the main data-plane application surfaces. [source]
Prefer service accounts over legacy API keys for new Atlas Admin API [source]
Remember the Atlas Admin API does not read or write your cluster data; [source]
it manages Atlas resources and access configuration. [source]
Design schemas around access patterns; data accessed together should [source]
generally live together. [source]
Prefer embedding when it lets common reads complete as a single-document [source]
fetch; use transactions only when requirements truly cross document [source]
Add indexes for repeated query patterns, but account for the write cost [source]
Use Performance Advisor, cluster metrics, alerts, and limits together; [source]
Atlas performance work is not just about query syntax. [source]
Treat **IP access lists, private networking, database users, and Atlas [source]
users** as distinct security controls with different scopes. [source]
Use Atlas Search for full-text/relevance workloads and **Atlas Vector [source]
Search** for semantic similarity/RAG workloads; they are related but not the [source]
Check deployment tier, MongoDB version, and Atlas limits before giving [source]
Know the Flex / Free / Dedicated tier model; Serverless and M2/M5 are [source]
Consider Atlas Stream Processing for event-driven and real-time ETL [source]
workloads before building external pipelines. [source]
For hybrid retrieval, use $rankFusion (MongoDB 8.1+) to merge [source]
full-text and vector results in a single query. [source]
Use Queryable Encryption for sensitive fields that still need to be [source]
queried; equality/range queries are GA, prefix/suffix/substring are in [source]

What Atlas expertise should mean

An Atlas-focused assistant should be able to reason across: [source]
control plane design: organizations, projects, deployments, users, [source]
network boundaries, backups, monitoring, limits, and automation [source]
application access: drivers, connection strings, connection pooling, [source]
query/index fit, aggregation, search, vector search [source]
operations: alerts, metrics, performance advisor, capacity, topology, and [source]
backup/disaster-recovery posture [source]
delivery models: UI, Admin API, CLI, Terraform, and Atlas Kubernetes [source]
streaming and event-driven: Atlas Stream Processing for real-time [source]
pipelines and time-series ingestion [source]
AI and agentic patterns: automated embedding, vector search, hybrid [source]
search, RAG, agent memory, and framework integrations [source]

Shared responsibility

Atlas Architecture Center explicitly frames Atlas through a **shared [source]
responsibility model**: MongoDB operates the underlying platform, while [source]
customers own their configuration, access control, and data policies. [source]
Atlas architectural guidance is organized around five pillars: [source]
**operational efficiency, security, reliability, performance, and cost [source]

Core control-plane hierarchy

Atlas operations are anchored around **organizations, projects, and [source]
deployments/clusters**. [source]
Many role assignments and operational actions are scoped differently at the [source]
organization vs project layer, so advice should always name the required [source]

Deployment and connectivity model

Atlas lets teams choose deployment tier (Free, Flex, or Dedicated), cloud [source]
provider, and region based on application latency, cost, and security [source]
requirements. Serverless instances and M2/M5 clusters are deprecated; Flex [source]
is their replacement. [source]
Application environments must satisfy both network access and [source]
database-user authentication to reach an Atlas deployment. [source]
Private connectivity choices include VPC/VNet peering and **private [source]
endpoints**, while public connectivity requires IP allow-listing. [source]

Atlas interaction methods inventory

This is the high-value inventory of how you can work with Atlas. For exact [source]
subcommands or endpoints, follow the linked source sections. [source]

Atlas Admin API standards

The Atlas Admin API is a REST-style control-plane API for Atlas resource [source]
The preferred authentication method is **OAuth2 service-account access [source]
tokens; HTTP Digest API keys are a legacy** option. [source]
Service-account tokens are obtained from [source]
POST https://cloud.mongodb.com/api/oauth/token and are reusable for about [source]
If your organization requires an API IP access list, token generation can [source]
happen from any IP, but API calls that use the token must come from an [source]
The Admin API does not expose application data access; reading and [source]
writing database documents still happens through cluster auth plus a driver or [source]
other data-plane client. [source]
As of March 2026, the Admin API enforces standardized rate limiting [source]
using a token-bucket algorithm. Automation code must handle `429 Too Many [source]
Requests responses and respect Retry-After` headers. [source]

Admin API categories to expect

For exact endpoints, use the API reference. In practice, Atlas Admin API work [source]
usually falls into these buckets: [source]
organizations and projects [source]
deployments / clusters / topology [source]
database users and access [source]
network access and IP lists [source]
alerts and monitoring metadata [source]
backup and restore administration [source]
inventory, IP address, and project metadata [source]

Atlas CLI standards

Atlas CLI is the terminal-native Atlas control-plane surface. [source]
atlas setup is the fast-start command that signs up or authenticates, [source]
creates a free database, loads sample data, adds the current IP to the access [source]
list, creates a database user, and connects with mongosh. [source]
Atlas CLI is appropriate for operator workflows and scripts, but when a [source]
task needs stable, reviewable automation, the Admin API or Terraform may be a [source]
Atlas docs explicitly surface CLI entry points for actions like **access list [source]
inspection** and other Atlas resource operations. [source]

Terraform

Use Terraform when you want declarative Atlas infrastructure management. [source]
The Atlas Terraform provider is the main Atlas IaC surface for [source]
provisioning/managing clusters and related Atlas resources from code. [source]
MongoDB’s guide expects service-account authentication for provider [source]
Provider 2.0 (2025) introduced semantic versioning with no-breaking-change [source]
guarantees in minor/patch releases. Projects on 1.x must follow the 2.0.0 [source]
Upgrade Guide before upgrading. See the refresh section below for details. [source]

Atlas Kubernetes Operator

Use Atlas Kubernetes Operator when Atlas resources need to be managed from a [source]
Kubernetes control plane. [source]
AKO manages Atlas state from custom resources like AtlasProject, [source]
AtlasDeployment, and AtlasDatabaseUser. [source]
AKO 2.0 changed deletion behavior: deleting a Kubernetes custom resource no [source]
longer deletes the Atlas resource by default. [source]
The docs explicitly warn to define desired config values explicitly to [source]
avoid inheriting Atlas defaults that can cause reconciliation loops. [source]

Network boundaries

Atlas only allows cluster client connections from the project’s **IP access [source]
list** unless you use private networking. [source]
IP access lists are project-wide, not per cluster. [source]
Atlas supports temporary IP access list entries with configurable [source]
For application connectivity, Atlas docs call out three main patterns: [source]
public IP allow-listing, VPC/VNet peering, and private endpoints. [source]

Identity boundaries

Atlas users are not the same as database users. [source]
Atlas users access the Atlas control plane; database users access MongoDB [source]
data-plane resources. [source]
Database users can be scoped with built-in roles, specific privileges, and [source]
Atlas supports multiple database-user auth methods, including SCRAM and [source]
X.509, with environment-sensitive guidance in the docs. [source]

Practical security defaults

Prefer private networking for higher-security production environments. [source]
Prefer service accounts for Atlas control-plane automation. [source]
Keep API and cluster network allow-lists narrow and time-bound where [source]
Use the minimum project/organization role that can perform the task. [source]
Keep the distinction between control-plane auth and data-plane auth [source]
explicit in design docs and code. [source]

Schema design

Model data around application access patterns. [source]
Keep data that is accessed together together. [source]
Use embedding when it reduces joins and multi-document coordination. [source]
Use referencing when data has a different access cadence, lifecycle, or [source]

Query and index design

Add indexes for repeated query shapes. [source]
Remember every additional index raises write cost. [source]
Atlas Performance Advisor is useful, but recommended indexes still need human [source]
judgment about workload frequency and write tradeoffs. [source]
Large arrays and $lookup-heavy designs are called out in Atlas docs as [source]
common sources of slow-query pain. [source]

Write and transaction design

Single-document operations are atomic; many practical designs should exploit [source]
that instead of defaulting to transactions. [source]
Include expected current state in update filters or use intent-specific [source]
operators like $inc where concurrency matters. [source]
Use transactions only when multi-document, multi-collection, or cross-shard [source]
atomicity is actually required. [source]

Driver usage

Production application code should usually target an **official MongoDB [source]
Atlas expertise includes choosing the right driver-level behavior for the [source]
language/runtime, not only knowing Atlas admin features. [source]

Atlas Search

Atlas Search is the embedded full-text and relevance-search system for Atlas. [source]
Search queries are expressed through aggregation pipeline stages such as [source]
$search and $searchMeta. [source]
Atlas Search supports **analyzers, autocomplete, pagination, faceting, [source]
scoring, static mappings, and dynamic mappings**. [source]
Use Atlas Search when the workload is fundamentally about **relevance-based [source]

Atlas Vector Search

Atlas Vector Search is the semantic/vector retrieval surface for Atlas. [source]
Use it for semantic search, hybrid search, and RAG/agentic retrieval [source]
Atlas docs call out ANN and ENN availability by MongoDB version; do [source]
not give vector-search guidance without checking version support. [source]
Automated Embedding uses the autoEmbed index field type with built-in [source]
Voyage AI models (voyage-4-large, voyage-4, voyage-4-lite, [source]
voyage-code-3) to generate embeddings on insert, update, and query without [source]
an external pipeline. Public preview as of May 2026. See the refresh section [source]
below for pricing and model details. [source]

Backup, reliability, and disaster-recovery standards

Atlas Cloud Backup uses the cloud provider’s native snapshot functionality and [source]
inherits provider redundancy guarantees. [source]
Atlas supports multi-region snapshot distribution for added redundancy and [source]
region-failure recovery posture. [source]
Backup restore/admin operations require the appropriate project backup [source]
roles; org-level access alone is not sufficient until explicitly added to the [source]
Topology matters: Atlas docs note restore caveats for sharded clusters after [source]

Alerts

Atlas alerts are configured around **conditions, thresholds, notification [source]
methods, and lifecycle management**. [source]
Alert policies can exist at the organization or project level. [source]
Atlas supports acknowledge/unacknowledge, disable/enable, and delete flows [source]

Metrics

Atlas collects metrics across servers, databases, and MongoDB processes. [source]
High-signal metrics called out in the docs include **connections, disk IOPS, [source]
disk usage, query targeting, and normalized system CPU**. [source]
Atlas exposes real-time metrics and Atlas Search metrics in addition [source]

Capacity and limits

Atlas limits around connections, shards, nodes, and topology are design [source]
inputs, not just operational trivia. [source]
Atlas docs explicitly recommend connection pooling, application tuning, [source]
autoscaling, and tier scaling when nearing connection limits. [source]

High-value method inventory

This is the condensed inventory to keep handy. For exhaustive method lists, use [source]
the linked references directly. [source]

Practical defaults for future Atlas coding/review tasks

Start with which Atlas surface the task belongs to: UI, Admin API, CLI, [source]
Terraform, AKO, or application driver. [source]
Separate Atlas control-plane actions from MongoDB data-plane actions [source]
before proposing code. [source]
For application design, start with access patterns, then schema, then [source]
indexes, then aggregation/search/vector shape. [source]
For operations, review alerts + metrics + limits + backup posture [source]
For security, review **network path + identity model + required role + [source]
auth method** together. [source]
For AI/search work, choose explicitly between Atlas Search, **Atlas [source]
Vector Search, or hybrid** patterns. [source]

Recent platform changes (2025-2026 refresh)

This section captures major Atlas and MongoDB platform changes since late 2024. [source]
Last refreshed: 2026-05-25. [source]

MongoDB server versions on Atlas

MongoDB 8.0 (GA October 2024): 32% query throughput improvement, 56% [source]
faster bulk writes, 200% faster time-series aggregations, 50x faster data [source]
distribution for sharding at 50% lower cost. Introduced default maximum time [source]
limits for queries and the ability to reject recurring problem queries. [source]
<https://www.mongodb.com/docs/manual/release-notes/8.0/> [source]
MongoDB 8.2 (2025): Public preview of enhanced Queryable Encryption [source]
(prefix, suffix, substring queries on encrypted fields), $currentDate in [source]
aggregate(), standardized spill-to-disk metrics in explain output. [source]
<https://www.mongodb.com/docs/manual/release-notes/8.2/> [source]
MongoDB 8.3 (May 2026): ~45% more reads and ~35% more writes vs 8.0, [source]
sub-100ms retrieval targets for agent workloads, new $hash and $hexHash [source]
aggregation expressions (MD5, SHA-256, XXH64), arrayIndexAs field in [source]
$map/$filter/$reduce, removeShard deprecated in favor of four new [source]
drain/removal commands, security hardening and native type-coercion [source]
<https://www.mongodb.com/docs/manual/release-notes/8.3/> [source]

Flex clusters (replaces Shared and Serverless tiers)

Atlas Flex tier is the unified replacement for M2, M5, and Serverless [source]
instances. It combines the best of Shared and Serverless into a single [source]
offering with dynamic scaling. [source]
As of March 2025, Serverless instances are no longer supported; existing [source]
instances were migrated to Free, Flex, or Dedicated clusters. [source]
As of May 2025, all M2/M5 clusters have been auto-migrated to Flex. [source]
As of January 2026, the old createGroupCluster (M2/M5) and [source]
createGroupServerlessInstance API endpoints only support Flex clusters. [source]
Flex includes 100 ops/sec and 5 GB storage by default, scales to 500 ops/sec [source]
dynamically, $8 base + usage-based billing capped at $30/month. [source]
<https://www.mongodb.com/docs/atlas/manage-flex-clusters/> [source]
Flex key limits: 500 connections max, 5 GB storage hard cap (no auto-expand), [source]
500 collections max, 100 databases max, MongoDB 8.0 minimum (auto-upgrade only). [source]
Flex does NOT support: Private Endpoints (no PrivateLink/VPC peering), [source]
Continuous Backup/PITR (daily snapshot only), BYOK encryption at rest, Database [source]
Auditing, Performance Advisor, Rolling index builds, allowDiskUse, server-side JS. [source]
Flex DOES support (unlike old M2/M5): Atlas Search, Atlas Vector Search, [source]
Change Streams, Triggers - but Vector Search shares resources with mongod on Flex; [source]
upgrade to M10+ with dedicated Search Nodes before production Vector Search. [source]
Migration is one-way: Flex → dedicated is supported (downtime required); [source]
dedicated → Flex downgrade is NOT supported. Download Flex snapshots before upgrading [source]
as they do not transfer to dedicated clusters. [source]
For full Flex decision matrix, pricing breakdown, and tooling migration, see [source]
mongodb-atlas-flex-serverless skill. [source]

Atlas Stream Processing (GA)

Atlas Stream Processing reached general availability as of March 2025. [source]
Enables continuous stream processing pipelines over Atlas data using [source]
aggregation-pipeline syntax. [source]
Supports emitting to Time Series Collections, Kafka headers, and [source]
multiple tiers (SP10 for low-traffic, SP30 for production). [source]
Available on AWS and Azure across global regions. [source]
<https://www.mongodb.com/docs/atlas/atlas-stream-processing/> [source]

Voyage AI acquisition and Automated Embedding

MongoDB acquired Voyage AI in February 2025 (~$220M) to embed [source]
high-accuracy embedding models directly into Atlas Vector Search. [source]
Automated Embedding (public preview May 2026) uses the autoEmbed index [source]
field type to automatically generate Voyage AI vector embeddings on insert, [source]
update, and query -- no external pipeline needed. [source]
Available models: voyage-4-large, voyage-4, voyage-4-lite, [source]
Pricing: per million tokens ($0.12 large / $0.06 standard / $0.02 lite); [source]
first 200M tokens free per account; Batch API gives 33% discount. [source]
<https://www.mongodb.com/docs/atlas/atlas-vector-search/> [source]

Lexical Prefilters for Vector Search

Lexical Prefilters allow advanced text and geo analysis filters (fuzzy [source]
search, phrase matching, wildcards, geoWithin) as prefilters before vector [source]
Unlike standard $vectorSearch filters (equals, range, exists), lexical [source]
prefilters use full analyzed-text capabilities from Atlas Search operators. [source]
Create a $search index with vector type fields and use [source]
$search.vectorSearch in aggregation pipelines. [source]
<https://www.mongodb.com/company/blog/product-release-announcements/semantic-power-lexical-precision-advanced-filtering-for-vector-search> [source]

Hybrid Search with `$rankFusion`

The $rankFusion aggregation operator merges and re-ranks results from [source]
multiple search pipelines (full-text + vector). [source]
Requires MongoDB 8.1+ on Atlas. [source]
Enables true hybrid search combining keyword precision with semantic [source]
intelligence in a single query. [source]
<https://www.mongodb.com/docs/atlas/atlas-vector-search/hybrid-search/> [source]

Search Nodes (dedicated search infrastructure)

Search Nodes are generally available on AWS, Google Cloud, and Azure for [source]
both development and production deployments. [source]
Provide dedicated infrastructure for Atlas Search and Vector Search, [source]
independent of database compute, with up to 60% query-time reduction. [source]
Multi-region Search Nodes are available in preview for multi-region and [source]
multi-cloud clusters. [source]
<https://www.mongodb.com/docs/atlas/atlas-search/> [source]

Queryable Encryption enhancements

Equality and range queries on encrypted fields are GA and production-ready [source]
at no additional cost on Atlas, Enterprise Advanced, and Community Edition. [source]
Prefix, suffix, and substring queries on encrypted string fields are in [source]
public preview starting MongoDB 8.2. [source]
<https://www.mongodb.com/docs/manual/core/queryable-encryption/> [source]

Terraform Provider 2.0

MongoDB Atlas Terraform Provider 2.0 shipped in 2025 with semantic [source]
versioning, no-breaking-change guarantees in minor/patch releases, eliminated [source]
hanging timeouts, and simplified advanced-cluster migrations. [source]
Migration required from 1.x; see the 2.0.0 Upgrade Guide. [source]
Atlas Architecture Center examples now target Provider 2.x. [source]
<https://www.mongodb.com/products/updates/terraform-mongodb-atlas-provider-2-0-now-available/> [source]

Atlas Admin API standardized rate limiting

Standardized rate limiting for the Atlas Admin API v2 became GA in [source]
March 2026, using a token-bucket algorithm. [source]
Automation and integration code should handle 429 Too Many Requests [source]
responses and respect Retry-After headers. [source]
<https://www.mongodb.com/company/blog/product-release-announcements/introducing-standardized-atlas-admin-api-rate-limiting> [source]

Atlas Architecture Center compliance additions

PCI DSS Compliance page added February 2026. [source]
HIPAA Compliance page added February 2026. [source]
Multi-region opinionated guidance, Reliability section, and Operational [source]
Readiness Checklist added August 2025. [source]
<https://www.mongodb.com/docs/atlas/architecture/current/changelog/> [source]

Atlas CLI updates (2025-2026)

atlas api subcommand reached GA in October 2025. [source]
TLS 1.3 support added for atlas api clusters commands (December 2025). [source]
atlas api aiModelRateLimits resetModelRateLimit command added April 2026. [source]
<https://www.mongodb.com/docs/atlas/cli/current/atlas-cli-changelog/> [source]

AI and agentic positioning

MongoDB is positioning Atlas as a converged datastore for agentic AI: [source]
operational data + vector search + stream processing + agent memory in one [source]
First-class integrations with LangGraph.js (long-term memory store, GA), [source]
and major agent frameworks. [source]
Atlas AI Integrations documentation consolidates RAG, agent, and embedding [source]
<https://www.mongodb.com/docs/atlas/ai-integrations/> [source]

Known ambiguities and guardrails

“All Atlas methods” is too large for a single static file. Use this context [source]
as the condensed expert map, then jump to the linked Admin API, CLI, [source]
driver, and Atlas feature references for exact syntax and complete inventories. [source]
Atlas docs are versioned and living. For vector search, API behavior, CLI [source]
commands, and limits, always confirm the current version or page timestamp [source]
when precision matters. [source]
Atlas advice often differs by **cluster tier, deployment topology, cloud [source]
provider, and MongoDB version**. Good answers should say which of those [source]

Cross-hub map — where every MongoDB topic lives

All MongoDB knowledge is split across four hubs (plus mongodb-kb for KB-article lookups and [source]
10gen for repo install/run). If a task's deep material is not in this hub's Sub-skill routing [source]
table, it is a reference file under a sibling hub - activate that hub or Read its references/<name>.md directly. [source]
High-overlap routing notes: [source]
Performance symptom triage (high CPU, cache pressure, slow queries, latency spikes) starts at atlas-diagnostics-expert, but storage-engine root-cause internals (WiredTiger cache fill / dirty trigger / eviction threads / reconciliation / checkpoints) are owned by mongodb-expert - cross-load mongodb-expert/references/mongodb-wiredtiger-internals.md (and mongodb-wiredtiger.md) for depth. [source]
Migration symptoms vs migration execution: live-cluster diagnosis → atlas-diagnostics-expert; the migration/mongosync runbook → mongodb-operations-expert. [source]
Atlas Search/Vector query syntax & index design → mongodb-atlas-expert; the slowness triage of a running search → atlas-diagnostics-expert. [source]

Children

Frontier under this node: MongoDB Atlas Data Federation

← the whole tree · 3D view· how to read this page

MongoDB Atlas

MongoDB Atlas Expert

Sub-skill routing table

Atlas platform and architecture

Atlas administration and automation

Connection, security, and access control

Data access and MongoDB application design

Atlas search and AI surfaces

Streaming, encryption, and newer surfaces

Operations, backup, observability, and limits

Atlas quick rules

What Atlas expertise should mean

Shared responsibility

Core control-plane hierarchy

Deployment and connectivity model

Atlas interaction methods inventory

Atlas Admin API standards

Admin API categories to expect

Atlas CLI standards

Terraform

Atlas Kubernetes Operator

Network boundaries

Identity boundaries

Practical security defaults

Schema design

Query and index design

Write and transaction design

Driver usage

Atlas Search

Atlas Vector Search

Backup, reliability, and disaster-recovery standards

Alerts

Metrics

Capacity and limits

High-value method inventory

Practical defaults for future Atlas coding/review tasks

Recent platform changes (2025-2026 refresh)

MongoDB server versions on Atlas

Flex clusters (replaces Shared and Serverless tiers)

Atlas Stream Processing (GA)

Voyage AI acquisition and Automated Embedding

Lexical Prefilters for Vector Search

Hybrid Search with `$rankFusion`

Search Nodes (dedicated search infrastructure)

Queryable Encryption enhancements

Terraform Provider 2.0

Atlas Admin API standardized rate limiting

Atlas Architecture Center compliance additions

Atlas CLI updates (2025-2026)

AI and agentic positioning

Known ambiguities and guardrails

See Also

Cross-hub map — where every MongoDB topic lives

Children