MuleSoft in Summer '26: API Catalog, MCP Servers, and Integration Intelligence

MuleSoft has always been Salesforce's answer to "how do I connect everything?" In Summer '26, that mission expands to AI agents and the open Model Context Protocol (MCP) — making MuleSoft a central control point for both APIs and agent tooling. Here's what integration teams should know.

API Catalog for Salesforce: one home for APIs and MCP servers

The big idea is the API Catalog for Salesforce — a single place to centralize and manage your APIs and MCP servers. With it you can:

• Bring in APIs from MuleSoft, Heroku, and Apex to create connections and activate actions for Salesforce automations.
• View, create, and activate Salesforce MCP servers, and sync MuleSoft MCP servers for use in Agentforce.

For anyone wrangling a sprawl of services, a unified catalog of both APIs and MCP servers is a genuine governance win.

Expose your agents and prompts to Claude and ChatGPT

This is the headline-grabber. Using API Catalog, you can:

• Map your agents to tools in custom Salesforce MCP servers — making your Agentforce agents available through Salesforce-hosted MCP servers to external clients like Claude or ChatGPT.
• Map prompt templates to MCP server tools — exposing your prompt templates the same way.
• Add annotations to MCP server tools (Read-only, Destructive, Idempotent, Open-world) to give clients hints about behavior.

In other words, MCP isn't just how Agentforce reaches out to other systems — it's also how other AI clients can safely reach your Salesforce capabilities. (For the inbound side of MCP, see our Agentforce platform guide.)

More MCP catalog capabilities (now GA)

• Bring MuleSoft MCP servers into API Catalog (GA) — sync MCP servers from Anypoint Exchange for use in Salesforce.
• Discover manually registered MCP servers (GA) — view servers you register in the Agentforce Registry or install from AgentExchange packages.
• View Named Query APIs in API Catalog and activate their actions (GA).

Integration Intelligence on Data 360

On the observability side, Integration Intelligence extends the value of your MuleSoft telemetry:

• Centralize metrics, logs, and traces in Salesforce Data 360 for long-term analytics.
• Gain visibility into historical trends and systemic patterns beyond real-time monitoring.
• Use the Integration Intelligence Lightning app with pre-made and custom Tableau Next dashboards for unified API and application insights.
• Use agent-assisted analysis to quickly find the cause of problems.
• Access these insights directly from Anypoint Platform with single sign-on.

Frequently asked questions

What is the API Catalog for Salesforce? A central place to manage APIs and MCP servers — bringing in APIs from MuleSoft, Heroku, and Apex, and creating, activating, and syncing MCP servers for use in Agentforce.

Can external tools like Claude or ChatGPT use my Salesforce agents? Yes. By mapping agents and prompt templates to tools in custom Salesforce MCP servers, you can make them available to external MCP clients like Claude or ChatGPT.

What are MCP tool annotations? Hints about a tool's behavior — Read-only, Destructive, Idempotent, or Open-world — that you add to tools in your custom Salesforce MCP servers.

What does Integration Intelligence do? It centralizes MuleSoft metrics, logs, and traces in Data 360 and surfaces long-term analytics through Tableau Next dashboards, with agent-assisted root-cause analysis.

The bottom line

MuleSoft in Summer '26 becomes the connective tissue of the agentic era: API Catalog manages both APIs and MCP servers, you can expose Agentforce agents and prompts to external AI clients, and Integration Intelligence brings deep observability into Data 360. If you're standardizing on MCP, MuleSoft is now a key part of the story.

See the inbound side of MCP in our Agentforce platform guide.