ServicesSalesforce consultingArchitectureImplementationOrg health checkDevelopmentIntegrations and APIsData migrationTesting and QAManaged supportSpecialist supportMarketing Cloud auditDeliverability
SolutionsMarketing Cloud EngagementMarketing Cloud NextSalesforce + BrazeAccount EngagementData Cloud (Data 360)Identity resolutionInsights and segmentsAgentforce and AIMCP integrationsSales and Service CloudReal-time personalisationIntegration solutionsEnterprise AI integrationBy industry
ArchitectureSalesforce CRMData 360Salesforce + BrazeAgentforceCustomer 360API-led integrationEvent-drivenMCP host, client and serverIdentity and consentMCP securityData governanceCI/CD and releasesOur public MCP server
ResourcesEngineering journalDecision guidesSolution blueprintsSalesforce productsProjects and achievementsSales ROI calculatorWhere we workChecklists and templatesPlanning toolsReadiness assessmentGlossaryAbout GeekMidstOur approachContact
CompanyAbout GeekMidstOur approachDelivery methodologyTrust and evidenceProjectsAchievementsBook a consultationContactWhere we workSecurity and data
Book a Consultation

MuleSoft and MCP

How MuleSoft's MCP Connector, MCP Bridge and Agent Fabric turn existing APIs into governed MCP tools, with each status as MuleSoft states it.

Type
Solution
Topic
MCP
Reading time
5 min
Updated
Author
GeekMidst Team
Reviewed by
Site owner
Banner for the GeekMidst solution “MuleSoft and MCP”, covering The business use case and What MuleSoft offers, and the stated status

Most organisations already have the APIs an AI agent needs: order status, customer lookup, inventory. The question is how to expose them as MCP tools without bypassing the policies, monitoring and ownership those APIs already have. MuleSoft offers several MCP capabilities for this. This page lists them with their stated status and explains how we would use them.

The business use case

  • Reuse governed APIs: expose a few existing APIs as MCP tools for Agentforce or other AI clients, instead of building new point-to-point code.
  • One control point: apply rate, security and data policies to MCP traffic in the same gateway as API traffic.
  • Inventory: know which MCP servers exist across the organisation, including ones teams set up themselves.

What MuleSoft offers, and the stated status

MuleSoft MCP capabilities, as documented

CapabilityWhat it doesStated status
Anypoint Connector for MCP (MCP Connector), v1.7.xA Mule app acts as an MCP server or client; Streamable HTTP recommended, SSE deprecated; Mule 4.9.6+ and Java 17Current connector; no stage label in our sources
MCP Bridge (Agent Fabric)Turns an API or SaaS system into an MCP server with no codeGA, MuleSoft '26 Mid-Year Release, 12 Aug 2026
API-to-MCP conversionConverting APIs to MCP"Currently in GA" as of 23 Jun 2025
Agent RegistryInventories agents, MCP servers and APIs, including a Trusted MCP cataloguePart of Agent Fabric; no separate stage stated
Agent ScannersDiscover MCP serversGA, 12 Aug 2026
Omni Gateway and Open PoliciesMCP, A2A and LLM policiesPart of Agent Fabric; no separate stage stated
Agent Fabric Context CatalogContext for agentsGA in Q3 2026, per the mid-year release
MuleSoft DX MCP ServerIDE use: design, deploy, assets; npm mulesoft-mcp-server; renamed 28 May 2026No status label in release notes
MuleSoft Platform MCP ServerRemote server for policies, discovery, cost and secrets; introduced 28 May 2026Status not stated

Recent Agent Fabric release notes add two details: from 31 August 2026, MCP Bridge exposes tool output schemas for Agentforce, and from 14 October 2026 remote MCP servers can register with OAuth 2.1 authorization code plus PKCE. The mid-year release also lists Informatica CDGC Metadata Search and Address Verification MCP servers as GA. Winter '27 materials mention AI Cost Management, Agent Kill Switch and MuleSoft Headless, with no status stated.

Choosing the build path

  • MCP Bridge when an existing API should become an MCP server quickly and its policies already fit. No code, but the API's design decides the tool's quality.
  • MCP Connector in a Mule app when a tool needs orchestration, such as calling two APIs, reshaping data or adding a confirmation token step.
  • A non-MuleSoft server only where neither fits; we would propose that per implementation.

Salesforce recommends an MCP gateway such as MuleSoft AI Gateway for cross-vendor governance. The layered pattern is in API-led integration architecture.

Proposed tool catalogue for an order API

This is a proposed design; we validate the APIs, edition and permissions for each implementation.

Example proposed tools over an existing order API

ToolPurposeRead/writeApproval needed
get_orderOrder header, lines and statusReadNo
list_invoicesInvoices for an account the caller may seeReadNo
check_stockAvailability for a product and locationReadNo
create_credit_noteRaise a credit note against an invoiceWriteYes, finance approver
cancel_orderCancel an orderWriteNot exposed; stays a human process

Authentication and access control

Keep backend credentials in MuleSoft, never in the AI client. For inbound calls, use OAuth: remote servers can register in Agent Fabric with OAuth 2.1 authorization code plus PKCE, and an MCP server must only accept tokens issued for it. For Agentforce, MuleSoft MCP servers appear in API Catalog after you enable Sync, create a connection (a managed named credential) and allowlist tools for the Asset Library.

Approvals for write tools

Write tools are marked with readOnlyHint: false and carry a clear description of their effect. In a Mule app, a write can be split into two steps: a preview that returns the exact change and a short-lived confirmation reference, and a commit that only accepts that reference after a person confirms in the host. Destructive operations are not exposed.

Privacy, limits and auditability

Return only the fields each tool needs; DataWeave mappings make that explicit. Apply rate-limit policies at the gateway and respect each backend's own limits; we do not assume numbers. Agent Registry gives an inventory of servers; keep logs of tool, caller, outcome and duration without storing payloads.

Errors, environments and testing

Map backend errors to tool results with isError: true and a plain reason, and keep protocol errors for unknown tools or bad parameters. Build in a non-production Anypoint environment, test with the MCP Inspector CLI, and promote through your usual pipeline. Order-to-cash patterns are in order-to-cash integration with MuleSoft.

Licensing

MCP Connector needs Mule 4.9.6 or later and Java 17. Agent Fabric capabilities and MuleSoft subscriptions vary; confirm entitlements with Salesforce or MuleSoft.

Implementation stages

  1. 01

    Inventory APIs

    List candidate APIs, owners and existing policies.

  2. 02

    Pick a path

    Choose MCP Bridge or MCP Connector per tool.

  3. 03

    Design tools

    Write narrow tool descriptions and output schemas; separate reads from writes.

  4. 04

    Apply policies

    Configure gateway policies and authentication.

  5. 05

    Test

    Test in non-production with the MCP Inspector CLI, including refusals and backend failures.

  6. 06

    Register and monitor

    Register in Agent Registry or API Catalog, allowlist tools and watch usage.

Available today vs proposed

ItemStatus
MCP Bridge, Agent ScannersGA, 12 Aug 2026 (MuleSoft)
MCP ConnectorCurrent connector, v1.7.x
MuleSoft Platform MCP ServerStatus not stated
GeekMidst order-API tool setProposed design

Questions we hear

Can MuleSoft turn an existing API into an MCP server?
Yes. MuleSoft states that MCP Bridge, GA from 12 August 2026, turns an API or SaaS system into an MCP server with no code.
Which transport should a Mule MCP server use?
MuleSoft recommends Streamable HTTP for the MCP Connector; SSE is deprecated.
Do we need Agent Fabric for MCP?
The MCP Connector's stated requirements are Mule 4.9.6 or later and Java 17. Agent Fabric adds registry, scanning, bridge and gateway governance; confirm entitlements with Salesforce or MuleSoft.
How do MuleSoft MCP servers reach Agentforce?
Through API Catalog: enable Sync, create a connection and allowlist tools for the Asset Library.

Turn your APIs into governed MCP tools

Tell us which APIs your agents need. We can choose between MCP Bridge and the MCP Connector and design the policies with you.

Book a consultation →

Tell us what you are trying to fix.

Share your Salesforce setup, goals and constraints. A practitioner reads every message and replies with questions, first thoughts or an honest view of whether you need outside help at all.

Talk to a Salesforce Expert
Book a Consultation