Quick Reference mode -- designed for live customer calls. Scroll to the objection you need. For deeper context, switch to any other tab.
What is MuleSoft Anypoint Platform?
One sentence
MuleSoft Anypoint Platform is an enterprise integration platform that connects any system, application, or data source -- cloud or on-premise -- using pre-built connectors, API management, and DataWeave transformation, enabling real-time data flow across the business without custom point-to-point code.
1,000+
pre-built connectors in Exchange
99.9%
uptime SLA (CloudHub 2.0)
4
deployment models (Cloud, Hybrid, On-prem, RTF)
Platform position: Anypoint is MuleSoft's full integration platform. IDP is one capability within it. When a customer has Anypoint, they have the runtime to build integrations, manage APIs, and process documents -- DataSkate configures and deploys on top of that.

Scope Boundary

In Scope (standard SOW)
  • System-to-system integration flows (agreed source and target pairs)
  • DataWeave transformation and field mapping
  • Error handling, retry logic, and dead letter queues
  • Scheduled and event-driven triggers
  • API design, deployment, and gateway policy setup
  • Monitoring dashboards and alerting configuration
  • Go-live testing, cutover support, and handoff docs
Out of Scope (needs separate SOW)
  • Integration flows beyond the agreed system list
  • Custom connector development for unsupported systems
  • Ongoing flow maintenance post-handoff (unless retainer)
  • Business logic changes to source or target systems
  • Portal or UI development for end-user interaction
  • Data cleansing or deduplication in source systems

Objection Handler

Click any question to expand the answer. These come from real customer calls.
"Why do we need a platform subscription? Can't you just write custom code to connect our systems?"
COMMERCIAL

Custom code is exactly what MuleSoft replaces. A point-to-point integration written in Python or Java works -- once. The moment a second system needs to talk to the same source, or the source system changes its API, the custom code breaks and someone has to fix it. You end up owning a fragile web of scripts with no visibility and no reuse.

Anypoint gives you managed connectors, centralized monitoring, and a reusable API layer. The platform subscription is the infrastructure cost that makes the implementation investment durable. Without it, every integration is a one-off that accrues maintenance debt from day one.

"We already have some integrations built internally. Why rebuild them in MuleSoft?"
COMMERCIAL

You do not have to rebuild everything at once. The right approach is to identify the highest-risk or highest-volume flows first -- the ones that break most often, or the ones where a failure has the most downstream impact -- and migrate those.

Existing custom integrations that are stable and low-maintenance can stay. The value of Anypoint compounds as more flows move onto a centralized, monitored, policy-governed platform. The goal is not a full rewrite; it is a shift in where new integrations are built so they do not add to the maintenance pile.

"What happens if MuleSoft has an outage? Our business stops."
RELIABILITY

CloudHub 2.0 carries a 99.9% uptime SLA backed by a multi-AZ AWS deployment. Historical uptime consistently exceeds that. But the more important answer is what happens when a transient issue does occur.

Anypoint integration patterns include retry policies, dead letter queues, and persistent object stores that ensure no message is lost during an outage. A brief runtime interruption does not result in lost transactions -- it results in a backlog that drains automatically when service resumes. For mission-critical flows, on-premise or hybrid deployment options remove the cloud dependency entirely.

"DataWeave looks proprietary. Are we locked in?"
TECHNICAL

DataWeave is MuleSoft's native transformation language -- yes, it is specific to the platform. The honest answer is that any enterprise integration platform creates some degree of lock-in because transformation logic lives inside the runtime.

The counter is that the integration logic itself is documented, version-controlled, and readable. If you ever moved off Anypoint, the mapping logic (what field maps to what, what transformations are applied) is fully documented in the DataWeave scripts. Rebuilding in another tool is a defined porting effort, not a black box you lose access to. This is true of any platform choice -- SSIS, Boomi, Workato all have the same dynamic.

"How long will this take? We need the integration live before our ERP go-live."
TIMELINE

A standard 2-3 system integration covering data sync, error handling, and go-live typically runs 8-12 weeks. Complexity scales with the number of flows, the quality of source system APIs, and how much data mapping needs to be designed from scratch versus following an existing canonical data model.

The highest-risk delays are: API documentation that does not match actual behavior, undefined field mapping decisions, and environment access issues (sandbox credentials, firewall rules, network permissions). Resolving these in the first two weeks of a project is the single biggest predictor of on-time delivery. We need a named technical contact on your side with access to the source and target systems from day one.

"Can Anypoint integrate with our legacy on-premise system that has no API?"
TECHNICAL

Yes. MuleSoft supports non-API integration patterns including database direct connections (JDBC for any SQL database), file-based exchange (SFTP, shared network drives), flat-file and EDI parsing, and SOAP/web services for older enterprise middleware. The Anypoint Runtime Fabric or on-premise Runtime Manager can run inside your network so the Mule runtime sits next to the legacy system without requiring it to expose a public API.

The approach depends on what the legacy system supports. During discovery, we document what interfaces exist (database tables, scheduled exports, batch files) and design the integration pattern accordingly. There is almost always a workable path -- the question is how much data extraction logic needs to live on the MuleSoft side.

MuleSoft Anypoint Platform

Anypoint is the unified integration and API management platform that connects enterprise systems -- cloud, on-premise, and hybrid -- using managed connectors, a native transformation language (DataWeave), and a centralized runtime. It replaces point-to-point custom code with governed, observable, reusable integrations.

Core Platform Capabilities

Anypoint Connectors
1,000+ pre-built connectors for Salesforce, NetSuite, SAP, Workday, ServiceNow, AWS, Azure, and hundreds more. REST, SOAP, JDBC, file, and messaging protocols all supported.
DataWeave
MuleSoft's native transformation language. Maps, filters, aggregates, and formats data between any source and target structure. Handles JSON, XML, CSV, EDI, and flat files natively.
API Manager
Design, publish, and govern APIs with rate limiting, authentication policies (OAuth, JWT, mTLS), SLA tiers, and client application access control. Flex Gateway extends policies to external APIs.
Runtime Manager
Central console for deploying, scaling, and monitoring Mule applications. Manages CloudHub 2.0 cloud workers, on-premise runtimes, and Runtime Fabric clusters from one interface.
Anypoint Exchange
Internal asset marketplace for reusable connectors, API specs, integration templates, and DataWeave scripts. Promotes reuse and reduces time-to-integration for subsequent projects.
Flex Gateway
Lightweight API gateway that applies Anypoint policies to APIs running outside the Mule runtime -- on-premise, in Kubernetes, or in third-party clouds. No runtime migration required.

Deployment Models

Anypoint supports four deployment models. Customers can mix models across flows within the same platform subscription. The runtime the Mule app runs on determines the model -- the design, monitoring, and API management tooling is always centralized in Anypoint.
Fully Managed Cloud
CloudHub 2.0
MuleSoft manages infrastructure, scaling, and patching. Multi-AZ deployment on AWS. Fastest to provision; 99.9% SLA. Best for net-new integrations without data residency constraints.
Container-Native
Runtime Fabric (RTF)
Mule runtime runs in your own Kubernetes cluster (on-prem or any cloud). You own the infrastructure; Anypoint manages the application layer. Best for data residency, network control, or existing K8s investment.
Traditional On-Premise
Hybrid / On-Premise Runtime
Mule runtime installed directly on your servers. Management plane stays in Anypoint (cloud) while the runtime and data stay on-premise. Common for legacy system access where network exposure is restricted.
API Policy Layer Only
Flex Gateway
Deploy as a lightweight sidecar or standalone proxy. Applies API Manager policies to existing APIs without moving workloads into Mule. Useful for securing third-party APIs or brownfield API environments.

Platform Reference Numbers

Standard Tier Capacity
Anypoint Integration Flows
30 Flows (Standard)
Message Volume
2M Messages / Year
API Gateway Calls
Metered by tier
Contract Term
12 Months (quarterly billing)

The Integration Lifecycle

01
Connect
Connector authenticates to source system and retrieves records via polling or event trigger
02
Transform
DataWeave maps source fields to target schema, handling type coercion and business rules
03
Route
Batch job chunks records for target write. Error handler catches failures and routes to DLQ
04
Monitor
Runtime Manager dashboards show message throughput, error rates, and SLA compliance in real time

Trigger Patterns

Every integration flow starts with a trigger. The pattern choice affects latency, error handling, and infrastructure requirements.
Sched­uled
Polling / Scheduled Trigger
A Scheduler component fires on a cron or fixed interval. The connector queries the source for new or changed records since the last run, tracked by a watermark timestamp in Object Store. Best for batch-style sync where near-real-time is acceptable.
Run every 15 min → query updated_at > last_run
Event
Event-Driven / Webhook Trigger
The source system pushes a notification to a Mule HTTP listener (webhook) or publishes to a message queue (Anypoint MQ, Kafka, SQS). Mule reacts immediately. Best for real-time requirements where latency matters.
Salesforce outbound message → Mule HTTP listener fires
File
File-Based / SFTP Drop
Mule polls a file location (SFTP, SharePoint, S3) for new files, processes each file, then moves or archives it. Common for ERP exports, EDI, and legacy system batch outputs that cannot expose an API.
SFTP poll every 5 min → parse CSV → upsert to target

DataWeave Transformation

What It Does
Field Mapping and Type Conversion
DataWeave is a functional transformation language. A mapping script reads the source payload and outputs the target schema. It handles format conversion (JSON to XML, CSV to JSON), date/time normalization, conditional logic, string manipulation, and array operations natively.
Why It Matters
All Mapping Logic Is Explicit and Auditable
Every field mapping is a readable script in version control. When a source system changes a field name or type, the DataWeave script is updated deliberately -- there is no hidden auto-mapping that silently breaks. The script is also the documentation: what field maps where is always readable without reverse-engineering a binary artifact.

Error Handling and Reliability

Every production Mule flow must have an Error Handler. A flow without one silently drops failed records. DataSkate includes error handling in all implementation work -- it is never optional.
1
On Error Continue
Catches the error, executes a recovery sub-flow (log, alert, write to DLQ), then continues processing the next record in the batch. The overall flow does not stop because one record failed.
2
Retry Policy
For transient errors (network timeout, rate limit), the connector retries with exponential backoff before escalating to the error handler. Configurable retry count and delay.
3
Dead Letter Queue (DLQ)
Records that exhaust retries and cannot be processed are written to a DLQ (Anypoint MQ, SQS, or a database table). The DLQ is monitored and reviewed -- failed records are never silently lost.
4
Alerting
Runtime Manager triggers alerts on flow failure or error rate thresholds. Alerts route to email, PagerDuty, Slack, or any webhook. On-call teams are notified before customers notice an impact.

Target Write Pattern (Upsert)

Standard Write Pattern
Upsert with External ID
Every write to the target uses upsert (insert-or-update) keyed on an external ID field that matches a unique identifier from the source system. This prevents duplicate records when the same object is synced more than once. The external ID field must be configured in the target system before the integration goes live.
Large Volumes
Batch Job Component
For high-volume integrations (thousands of records per run), the Mule Batch Job component splits the payload into configurable chunk sizes (typically 200 records), processes each chunk in parallel, and handles per-chunk error recovery independently. Prevents memory exhaustion and timeout failures common in large single-payload writes.
These are the questions customers ask during proposal and scoping calls -- organized by category. Expand any question to see a full answer you can adapt in the moment.

Technical

Architecture & Functionality
What systems can Anypoint connect to?

Any system with a network-accessible interface. Anypoint Exchange has 1,000+ pre-built connectors covering Salesforce, NetSuite, SAP, Workday, ServiceNow, AWS, Azure, Google Cloud, Shopify, HubSpot, Snowflake, Oracle ERP, and hundreds more. For systems without a connector, MuleSoft supports REST, SOAP, JDBC, SFTP, SMTP, and raw socket protocols -- building a custom connector is rarely needed because one of these protocols usually covers the gap.

Does MuleSoft store our data? Where does data actually go?

Anypoint is a data-in-transit platform, not a data store. Data moves through the Mule runtime -- source to target -- without being persisted in MuleSoft's infrastructure, unless a specific feature like Anypoint MQ (message queue) or Object Store (watermark/state) is used. Even then, those are ephemeral stores for operational state, not a data warehouse.

For CloudHub 2.0 deployments, the runtime runs in AWS. You can configure which AWS region (us-east-1, eu-west-1, etc.) to meet data residency requirements. For fully air-gapped data requirements, Runtime Fabric runs the Mule runtime entirely within your own infrastructure.

How does Anypoint handle API versioning?

API Manager supports versioned API instances. When a breaking change is introduced, a new API version is published (v2) alongside the existing one (v1). Consumers are migrated on a defined schedule; the old version is deprecated with a sunset date. This decouples the backend change from consumer readiness and prevents breaking live integrations.

Anypoint Exchange maintains a version catalog for all published APIs, so consumers always know what version they are contracted against and when it is being retired.

Can Anypoint handle real-time vs. batch? Which should we use?

Both patterns are supported. The right choice depends on use case:

Use real-time (event-driven) when latency matters to the business -- a new order in ERP needs to appear in the CRM within seconds, or a customer service ticket should trigger an immediate notification. Uses webhooks, Anypoint MQ, Kafka, or SQS.

Use batch (scheduled polling) when latency tolerance is 5-60 minutes and volumes are high. Simpler to operate, easier to recover after failures, and cheaper to run at scale. Uses Scheduler + watermark + Batch Job.

Most integration landscapes use a mix. Real-time for customer-facing workflows, batch for reporting, reconciliation, and back-office sync.

What if a target system's API is rate-limited?

Rate limiting is a common constraint with Salesforce (API call limits), NetSuite, and SaaS platforms. MuleSoft handles this through three mechanisms: (1) Batch Job chunking -- rather than sending 10,000 individual API calls, batch them into 200-record chunks and pace the throughput; (2) Retry with exponential backoff -- on a 429 rate limit response, the connector waits and retries automatically; (3) API Manager throttling policies -- apply rate limits at the Anypoint API layer to prevent your own integrations from overwhelming a downstream system.

During scoping, we document the rate limits of each target system and design the flow to stay within them. Salesforce's API call limits are the most common constraint and are accounted for in standard design patterns.

Commercial & Scope

Budget & Engagement
What does a standard integration implementation include?

Standard scope: source connector configuration, DataWeave field mapping, error handling with DLQ, retry policies, watermark-based incremental sync, upsert write to target with external ID, Runtime Manager monitoring setup, alerting, go-live cutover support, and handoff documentation including mapping specs and operational runbook.

Out of scope by default: custom connector development, source system data cleanup, business logic changes in the source or target, ongoing maintenance post-handoff (handled via retainer), and flows beyond the agreed system list.

How is the implementation priced -- fixed or time and materials?

DataSkate scopes integration projects as fixed-fee SOWs when the system list and field mappings are defined. The SOW specifies the number of flows, the source/target pairs, and the data objects in scope. Changes beyond that list are change orders.

When discovery is incomplete (field mapping not defined, source system API behavior unknown), we recommend a short paid discovery engagement first to reduce scope risk before a fixed-fee SOW is appropriate.

How long does implementation take?

Standard 2-3 system integration: 8-12 weeks. Drivers of timeline variance: number of integration flows, whether source systems have documented APIs, how much field mapping is pre-defined vs. needs to be designed, and how quickly the client can provision sandbox access and review mapping specs.

The fastest projects share one characteristic: a named technical point of contact on the client side with authority over integration design decisions who responds to questions within 24 hours.

Compliance & Security

Risk & Governance
Is MuleSoft HIPAA-compliant?

Salesforce (MuleSoft's parent company) offers HIPAA-eligible services and will sign a Business Associate Agreement (BAA). The platform supports encryption in transit (TLS 1.2+) and at rest, audit logging, access controls, and role-based permissions that satisfy HIPAA technical safeguard requirements.

Compliance requires more than platform capability -- it requires how you configure it. The integration design must avoid persisting PHI unnecessarily, mask sensitive fields in logs, and enforce credential management through Anypoint Secrets Manager. DataSkate builds HIPAA-aware integration patterns as part of our standard healthcare implementation approach.

How does Anypoint handle credentials and secrets?

Credentials are stored in Anypoint Secrets Manager -- an encrypted vault integrated directly into the Mule runtime. Connectors reference a secret alias, not the actual credential. Secrets are never stored in flow configuration files, version control, or CloudHub properties in plaintext.

For on-premise deployments, credentials can stay entirely within the customer's network. The Mule runtime reads secrets from the local Secrets Manager vault or can integrate with enterprise secret stores like HashiCorp Vault or AWS Secrets Manager.

What audit and logging does Anypoint provide?

Runtime Manager provides flow-level execution logs with configurable log levels (INFO, DEBUG, ERROR). Anypoint Monitoring (included in higher tiers) adds application performance metrics, distributed tracing, and custom dashboards.

For compliance use cases, logs can be forwarded to external SIEM systems (Splunk, Datadog, CloudWatch) via log forwarding configuration. Every API call through API Manager is logged with client ID, timestamp, endpoint, and response code -- this is the audit trail for API governance and access reviews.

Post-Go-Live

Operations & Support
What happens when a source system changes their API after go-live?

Minor changes (new optional fields, endpoint URL changes, token refresh updates) are typically a small DataWeave or connector config update -- usually addressed within a few days on a retainer, or as a small change order off retainer.

Breaking changes (deprecated endpoints, schema overhaul, authentication method change) require a more substantial update engagement. The DLQ is your early warning system: a spike in failed records after a source system update is the first signal that something changed on the API.

We recommend a post-go-live retainer that includes proactive monitoring and response to API change notifications from source systems. Most SaaS vendors publish API deprecation timelines 3-6 months in advance; subscribing to their developer changelogs is part of the operational playbook we hand off at go-live.

Can your team take on ongoing support after the implementation?

Yes. DataSkate offers post-go-live retainer engagements covering: DLQ monitoring and remediation, flow adjustments for API changes, new object or field additions within an existing integration, performance tuning, and Runtime Manager alert response.

Retainers are scoped as monthly hour blocks. The right retainer size depends on how many integrations are live and the expected rate of change in the source systems. We typically recommend starting with a 10-20 hour monthly block for a 2-3 system integration and adjusting after the first 60 days in production.

Anypoint's security model is layered -- platform-level controls from Salesforce/MuleSoft, policy-level controls in API Manager, and implementation-level controls built into each integration flow. All three layers are required for a compliant production deployment.

Platform Security Controls

Anypoint Secrets Manager
All credentials, API keys, certificates, and tokens are stored in an encrypted vault. The Mule runtime resolves secrets at deploy time -- credentials never appear in flow configuration or version control.
Encryption in Transit
All Mule-to-system communication uses TLS 1.2 or higher. Mutual TLS (mTLS) is available for service-to-service authentication where certificate pinning is required. Plaintext HTTP is blocked by platform policy in production environments.
Role-Based Access (RBAC)
Anypoint Control Plane uses fine-grained roles: Organization Admin, Environment Admin, Developer, Viewer, and custom role combinations. SSO via SAML 2.0 or OIDC integrates with enterprise identity providers (Okta, Azure AD, Ping).
Audit Logging
All control plane actions (deployments, policy changes, environment access) are logged in the Anypoint Audit Log. Logs are tamper-evident, retained per Salesforce's data retention policy, and exportable to external SIEM systems.
Network Isolation
CloudHub 2.0 supports Private Spaces -- dedicated VPCs with private load balancers, IP allowlisting, and no public internet exposure. Ideal for integrations that must not traverse the public internet between source and runtime.
On-Premise / Hybrid Options
Runtime Fabric and on-premise runtime options allow the Mule application to run entirely within the customer's network. Only control plane metadata (deployment status, metrics) traverses to MuleSoft's cloud. Data never leaves the customer's environment.

API-Level Security (API Manager)

API Manager applies policies to APIs managed by Anypoint. Policies enforce access control, rate limiting, and threat protection without changing the underlying Mule flow code.
Auth
OAuth 2.0 / JWT Validation
Enforce token-based authentication on every inbound API call. Validate JWT signature, expiry, and claims before the request reaches the Mule flow. Supports client credentials, authorization code, and implicit grant types.
Rate
Rate Limiting and SLA Tiers
Assign API consumers to SLA tiers (e.g., free: 100 calls/min, enterprise: 5,000 calls/min). Rate limit enforcement happens in the API gateway layer before any processing cost is incurred. Quota breaches return 429 without hitting the Mule runtime.
IP
IP Allowlist / Denylist
Restrict API access to specific IP ranges. Common for partner APIs that should only be callable from known partner infrastructure, or internal APIs that should never be reachable from the public internet.
Threat
JSON / XML Threat Protection
Validates inbound payloads against structural limits (max depth, max array size, max string length) to prevent injection and denial-of-service attacks through malformed payloads. Applied before deserialization.

Certifications and Compliance Posture

Certification / Standard Status Notes
SOC 2 Type II CertifiedAnnual audit by third-party auditor. Report available under NDA.
ISO 27001 CertifiedCovers MuleSoft/Salesforce cloud infrastructure and operations.
HIPAA Eligible BAA AvailableSalesforce signs BAA for HIPAA-covered workloads. Requires appropriate CloudHub region and configuration.
PCI DSS PCI-ReadyCardholder data flows can be scoped out of the Mule runtime. Architecture review required for PCI scope reduction.
GDPR SupportedEU data residency available (CloudHub EU region). DPA available from Salesforce. Data minimization principles must be applied in flow design.
FedRAMP Moderate (Government Cloud)Available via Salesforce Government Cloud Plus. Requires separate licensing.
These are real constraints to be aware of -- not showstoppers, but things that affect scoping decisions, timelines, and design choices. Know them before the customer asks.
DataWeave Learning Curve
DataWeave is a powerful but opinionated functional language. Developers familiar with JavaScript, Python, or Java need 2-4 weeks to become productive. Complex transformations (deeply nested JSON, multi-input joins, conditional aggregations) take additional time to get right and test thoroughly.
Mitigation: DataSkate developers are certified. Client developers who need to maintain flows should be budgeted for a DataWeave training engagement as part of the SOW.
Licensing Complexity at Scale
Anypoint licensing is based on a combination of integration flows, message volume, and API calls. As integrations scale, customers can hit tier limits unexpectedly mid-contract. The metered nature of some tiers can make cost forecasting challenging for high-growth environments.
Mitigation: During scoping, we model projected message volumes and API call counts to right-size the initial license tier. We flag flows that are likely to generate high volume so they are accounted for in the contract.
CloudHub 2.0 Cold Start Latency
Applications deployed on shared CloudHub workers experience a startup delay (cold start) when a worker has been idle and is restarted. Cold starts can add 10-30 seconds of unavailability. Not a concern for continuous-traffic integrations, but relevant for low-frequency scheduled flows that must fire immediately on schedule.
Mitigation: Use dedicated workers (higher cost but eliminates cold starts), or design flows to tolerate a brief startup delay. Runtime Fabric deployments on persistent K8s pods do not have this issue.
Connector Limitations for Niche Systems
While the connector library is vast, niche ERP systems, legacy middleware, and industry-specific platforms may not have a pre-built connector. In those cases, integration requires building against a REST, SOAP, or JDBC interface directly -- which is always possible but takes longer than using a certified connector and requires more testing.
Mitigation: During discovery, we inventory connector availability for every system on the integration map. Systems without a connector get explicit time estimates for the custom connection layer.
Not a Data Warehouse or ETL Tool
Anypoint moves data between operational systems in real-time or near-real-time. It is not designed for large-scale analytical ETL, historical data loads of millions of rows, or data warehouse ingestion. Attempting to use it for bulk historical migration creates performance and cost problems.
Mitigation: Historical data migration is handled separately (direct database load, Salesforce Data Loader, or a dedicated ETL tool like Fivetran). Anypoint takes over for ongoing operational sync after the initial load is complete.
Debugging Complexity in Long Flow Chains
Long Mule flows with many components, sub-flows, and error handlers can be difficult to debug in production when issues arise. Log verbosity must be balanced -- too little and failures are opaque, too much and log costs and noise increase. Distributed tracing is available in Anypoint Monitoring (Titanium tier) but not in base subscriptions.
Mitigation: DataSkate builds structured logging into every flow (correlation IDs, checkpoint logs at key steps). This is part of our standard implementation pattern and is explicitly called out in the operational runbook delivered at go-live.
Anypoint Platform
MuleSoft's unified integration and API management platform. The umbrella product that contains all tools -- connectors, DataWeave, API Manager, Runtime Manager, Exchange, and Flex Gateway.
API Manager
The Anypoint console for designing, publishing, and governing APIs. Applies policies (auth, rate limiting, threat protection) to APIs without changing the underlying Mule flow code.
Anypoint MQ
MuleSoft's managed message queue service. Used for async communication between Mule apps, or as a dead letter queue for failed records. Persists messages until consumed, enabling reliable delivery without tight coupling.
Batch Job
A Mule component that splits a large payload into configurable chunks (e.g., 200 records) and processes each chunk in parallel. Handles per-chunk error recovery independently. Essential for high-volume integrations.
CloudHub 2.0
MuleSoft's fully managed cloud deployment option. Runs Mule applications on AWS infrastructure across multiple availability zones. Carries a 99.9% uptime SLA. MuleSoft manages patching, scaling, and infrastructure.
Connector
A pre-built integration adapter that handles authentication, API communication, and data retrieval for a specific system (Salesforce, NetSuite, SAP, etc.). Encapsulates the protocol details so flow developers work at the logical level.
DataWeaveDW
MuleSoft's native data transformation language. Used to map source fields to target schemas, convert formats (JSON, XML, CSV, EDI), apply business rules, and filter or aggregate data. All mapping logic is a readable script in version control.
Dead Letter QueueDLQ
The destination for records that fail processing after all retries are exhausted. The DLQ is monitored and reviewed -- failed records are never silently lost. Implemented using Anypoint MQ, SQS, or a database table depending on the architecture.
Error Handler
A required Mule component that catches and handles exceptions in a flow. "On Error Continue" processes the error and continues the batch. "On Error Propagate" stops the flow and escalates. Every production flow must have one.
Anypoint Exchange
The internal marketplace for reusable integration assets: connectors, API specs (RAML/OAS), integration templates, and DataWeave libraries. Promotes reuse across projects and reduces time-to-integration for repeat patterns.
External ID
A field in the target system (e.g., Salesforce custom field) that stores the unique identifier from the source system. Used as the key for upsert operations to prevent duplicate records when the same object is synced multiple times.
Flex Gateway
A lightweight API gateway that applies Anypoint API Manager policies to APIs running outside the Mule runtime -- on Kubernetes, on-premise servers, or in other clouds. Does not require migrating the API into a Mule flow.
Mule Runtime
The Java-based engine that executes Mule applications. Can be deployed to CloudHub 2.0 (managed), Runtime Fabric (containerized), or on-premise servers (customer-managed). All three use the same Mule application artifact.
Object Store
A key-value store built into the Mule runtime. Primarily used to persist the watermark timestamp for incremental polling flows -- the "last run" timestamp is stored here so the next poll only retrieves records changed since then.
RAML / OAS
API specification formats used in Anypoint Exchange. RAML (RESTful API Modeling Language) is MuleSoft's native format; OAS (OpenAPI Specification) is the industry standard. API specs are the contract that consumers build against before the implementation is built.
Runtime FabricRTF
MuleSoft's container-native deployment option. Runs Mule applications in the customer's own Kubernetes cluster (on-prem or any cloud). The customer owns the infrastructure; Anypoint manages the application lifecycle. Provides data residency and network control not possible with CloudHub.
Runtime Manager
The Anypoint console for deploying, scaling, and monitoring Mule applications. Shows app status, memory/CPU usage, log streams, and alert configuration. Used by both DataSkate and the client's ops team post-go-live.
Secrets Manager
Anypoint's encrypted credential vault. Stores API keys, passwords, certificates, and OAuth tokens. Mule flows reference secrets by alias -- the actual value is never exposed in configuration files, version control, or deployment properties.
Scheduler
A Mule source component that triggers a flow on a fixed interval or cron expression. Used as the starting point for polling-based integrations. Works with Object Store watermark to enable incremental (delta) sync rather than full re-sync on every run.
Upsert
A write operation that inserts a record if it does not exist, or updates it if it does -- based on a matching key (external ID). The standard pattern for all target writes in Anypoint integrations to prevent duplicate records from incremental sync runs.
Watermark
A timestamp (or sequence number) persisted in Object Store that marks the point at which the last successful poll completed. The next poll queries only for records modified after the watermark, enabling efficient incremental sync without re-processing the full dataset on every run.
Hover over any component to see a detailed explanation. This diagram shows a standard Anypoint integration architecture with polling and event-driven trigger patterns.
Source Systems (trigger side)
ERP (NetSuite / SAP)
CRM (Salesforce)
SFTP / File Drop
External REST API
Legacy Database (JDBC)
Connector (polling or event-driven) -- authenticated, managed, retry-aware
ANYPOINT PLATFORM -- MULE RUNTIME
Trigger Layer
Scheduler / HTTP Listener / MQ Consumer
Scheduler (cron)
HTTP Listener
MQ Consumer
Connector + Retrieval
Query / Read with Watermark
Connector Query
Object Store Watermark
Filter (delta only)
DataWeave Transformation
Map / Filter / Aggregate / Format Convert
Field Mapping
Type Coercion
Business Rules
Format Convert
Batch Job (High-Volume)
Chunk + Parallel + Per-Record Retry
Split into 200-record chunks
Parallel processing
Failed records → DLQ
Error Handler
Retry Policy + DLQ + Alert
Exponential backoff retry
Write to DLQ
Alert (email / PagerDuty)
Target Connector (upsert with external ID)
Target Systems (write side)
Salesforce
ERP / NetSuite / SAP
Workday / HCM
Snowflake / Data Platform

Key Design Decisions

Trigger Pattern
Polling vs. Event
Choose polling (Scheduler + watermark) for batch tolerance. Choose event-driven (HTTP/MQ) for real-time requirements. Most landscapes need both.
Write Pattern
Always Upsert
Upsert with external ID prevents duplicates when the same record is processed multiple times. Never use insert-only in an integration that runs repeatedly.
Volume Handling
Batch for Scale
Flows processing >500 records per run should use the Batch Job component. Single large payloads time out and consume excessive memory in the runtime.
Error Strategy
DLQ is Mandatory
Every flow that writes to a system must have a DLQ path. Silent failure is never acceptable. Failed records must be captured, monitored, and replayable.