Salesforce Integration Solutions
An overview of how we connect Salesforce to other systems: patterns, platforms, Braze and Data 360 paths, and how failures are caught and fixed.
Salesforce rarely works alone. Orders come from ERP, events from websites and apps, engagement data from marketing platforms, and now AI tools want governed access to records and actions. Integration decides whether these connections are reliable or fragile.
This page is an overview of our integration solutions: the patterns, the platforms, how Braze and Data 360 fit, and how we handle errors and monitoring. Each section links to the page that goes deeper. For delivery, see Salesforce integration services and MuleSoft integration services.
The business problem
Point-to-point integrations accumulate. Each one works until a field changes, a token expires or volume spikes, and then failures are found by users rather than monitoring. The goal is a small set of well-understood patterns, each with clear ownership, error handling and observability.
Integration patterns
Patterns and when to use them
| Pattern | What it is | Use when |
|---|---|---|
| API-led | Layered APIs (system, process, experience) that hide back-end detail | Several consumers need the same data; systems change independently |
| Event-driven | Platform Events and Change Data Capture published to subscribers | Changes must reach other systems quickly without tight coupling |
| Batch or ETL | Scheduled bulk movement and transformation | Large volumes where minutes or hours of delay is acceptable |
| Data virtualisation | Salesforce Connect external objects or Data 360 zero copy | Data should stay in its source system |
| MCP for AI tools | Governed tool access for AI clients and agents | AI assistants or agents need records and actions with user permissions |
API-led integration
API-led integration separates system APIs from process and experience APIs, so a change to a back-end system does not break every consumer. See the API-led integration architecture and middleware vs point-to-point APIs.
Event-driven integration
Platform Events and Change Data Capture let Salesforce publish changes. External clients subscribe through the Pub/Sub API (gRPC, HTTP/2, Avro); Apex triggers subscribe inside Salesforce and run asynchronously after commit. High-volume platform events and change events are retained for 72 hours, so subscribers can replay after short outages. Salesforce labels PushTopic and generic events "(Legacy)". See the event-driven architecture and Platform Events for decoupled integrations.
Batch and ETL
Batch suits nightly loads, migrations and reconciliation. Design for idempotency, so a re-run does not create duplicates, and for partial failure, so one bad record does not stop the batch.
MCP for AI tools
The Model Context Protocol gives AI clients a standard way to call tools. Salesforce Hosted MCP Servers became generally available on 29 April 2026 for Enterprise Edition and above; the docs say they are "intended only for customers with Flex Credits". Each call runs as the authenticated user through an External Client App with OAuth and PKCE. See MCP integrations and MCP vs direct API integration.
MuleSoft
MuleSoft's Anypoint Platform provides API design, integration and management. MuleSoft Agent Fabric adds an agent registry, AI Gateway and governance for agents and MCP servers. MCP Bridge, which turns an API or SaaS system into an MCP server without code, is GA per MuleSoft's mid-year release of 12 August 2026. The Anypoint Connector for MCP lets a Mule app act as an MCP server or client. MuleSoft Composer is still documented for simpler, no-code flows. Check MuleSoft licensing with your account team. See MuleSoft and MCP and our order-to-cash integration with MuleSoft blueprint.
Braze integration
We found no native, vendor-supported connector between Salesforce and Braze in official documentation. Braze documents a community-submitted Sales Cloud pattern that uses Braze webhooks to create and update Leads, one way only. Production designs usually put middleware between the two, using Change Data Capture or Platform Events outbound from Salesforce and the Braze REST API or Cloud Data Ingestion inbound to Braze. Salesforce has announced an agreement to acquire Braze; until it closes they are separate companies, and no joint integration has been announced. See Salesforce and Braze.
Data 360 connectors
Data 360 offers "zero cost connectors" for Sales, Service, Industries, B2C Commerce and Marketing Cloud Engagement, plus file, API, SDK and zero copy paths. Outbound, segments go to activation targets and data actions reach Platform Events, Marketing Cloud or a webhook. See Data 360 activation.
Security boundaries
Use named credentials and external credentials rather than credentials in code; see named credentials explained. Salesforce has announced that connected apps support ends in Summer '27, with migration to External Client Apps, so new integrations should start there. Give each integration its own least-privilege user.
Error handling and monitoring
Failure handling
| Failure | Response |
|---|---|
| Target system unavailable | Retry with backoff; queue messages; alert after a threshold |
| Rate limit reached | Respect the provider's limits and retry headers; spread load |
| Bad record | Route to a dead-letter queue with the error; continue the batch |
| Expired or revoked token | Alert immediately; never fall back to shared credentials |
| Missed events | Replay from the retention window; reconcile with a periodic batch check |
Monitor message counts, error rates, latency and queue depth per integration, with alerts routed to a named owner. Our integration error handling checklist and API error handling guide cover the detail.
Deployment model
Integrations run in your Salesforce org, your MuleSoft or iPaaS tenant, or your cloud. GeekMidst works remotely from India and builds in sandboxes first.
Implementation stages
- 01
Assess
Inventory systems, data flows, volumes and current failures.
- 02
Choose patterns
Pick a pattern per flow and agree ownership.
- 03
Design
Define contracts, security, error handling and monitoring.
- 04
Build and test
Build in sandboxes; test failure paths, not just success.
- 05
Operate
Monitor, alert and review on a set cadence.
Testing strategy
Test each integration under expected peak volume, with the target system unavailable, with a malformed record and with an expired token. Confirm that replay and reconciliation recover cleanly.
Prerequisites and limitations
Each platform has its own licences and limits: Salesforce API limits, MuleSoft subscriptions, Flex Credits for hosted MCP servers and each vendor's rate limits. We check limits in each vendor's documentation rather than assuming them.
Related
Integration complexity assessment
Score your integration landscape
→Securing API integrations
Authentication and least privilege
→Integration whitepaper
Patterns in more depth
→Enterprise AI integration
Connecting Salesforce to AI safely
→Questions we hear
Do we need MuleSoft?
Is there a native Salesforce–Braze connector?
What should replace connected apps?
Should AI tools use MCP or direct APIs?
Sources
- https://developer.salesforce.com/docs/atlas.en-us.change_data_capture.meta/change_data_capture/cdc_subscribe.htm
- https://developer.salesforce.com/docs/atlas.en-us.api_streaming.meta/api_streaming/using_streaming_api_durability.htm
- https://developer.salesforce.com/blogs/2026/04/salesforce-hosted-mcp-servers-are-now-generally-available
- https://blogs.mulesoft.com/news/mulesoft-26-mid-year-release-create-ai-experiences/
- https://www.braze.com/docs/user_guide/get_started/b2b_use_cases/b2b_salesforce_sales_cloud
- https://www.mulesoft.com/
Plan your integration landscape
Tell us which systems need to connect, and we will propose patterns and a first phase.
