An Overview of Salesforce MCP Servers and Connectors

Top 10 Salesforce MCP Connectors

Salesforce’s adoption of the Model Context Protocol (MCP) – an open standard for AI-to-system integration – has spurred a wave of connectors that let AI tools (like ChatGPT or Claude) query and manage Salesforce data. In 2025 Salesforce even launched its own pilot Hosted MCP Servers so AI agents can “securely access your Salesforce data” without custom code ^[1]. Below are 10 prominent MCP connectors and servers for Salesforce, with Cirra.ai’s solution listed first as requested.

1. Cirra AI – Salesforce Admin MCP Server

Cirra AI claims the first commercial MCP server built specifically for Salesforce administration ^[2]. This “ Salesforce Admin MCP” server lets admins use plain-English AI commands to manage org setup and metadata instead of clicking through the UI. As Cirra puts it, the MCP integration “ revolutionizes Salesforce administration by bringing AI reasoning and execution directly into admin workflows” ^[2]. In practice, an administrator can ask an AI (in Claude, ChatGPT, etc.) to perform org changes (create fields, update layouts, configure objects) and watch the server orchestrate the Salesforce API calls behind the scenes. Cirra’s offering is aimed at Salesforce admins who want to “harness the full power of AI” for routine org tasks ^[2].

2. Coupler.io – Salesforce MCP Connector

Coupler.io provides a no-code MCP integration that links Salesforce data to AI agents. Its Salesforce MCP connector is marketed as an “AI-Powered CRM data analyst” where users can “connect Salesforce with AI and get instant answers” about pipeline, trends, and reports via natural language ^[3]. Essentially, Coupler flows pull Salesforce objects (Accounts, Contacts, Leads, Opportunities, custom reports, etc.) and feed an MCP server, so you can ask the AI questions like “Which leads in Q1 have >50% win probability?”. Coupler’s interface guides users through connecting Salesforce and then simply typing queries. This connector abstracts away the setup, so non-technical users can chat with Salesforce data. The company highlights use cases like spotting high-value deals, generating segment reports, and tracking conversion rates by querying Salesforce “through simple conversations” ^[3].

3. SyncHub – Plug-and-Play Salesforce MCP Connector

SyncHub offers a plug-and-play MCP server that lets you “chat with Salesforce” via any AI interface. Their Salesforce connector runs in the background, pulling real-time data and modeling it for AI queries. SyncHub emphasizes context: its server “can accept your natural language question and return a dataset that crosses multiple endpoints (or even multiple cloud services – we support 70+ connectors in addition to Salesforce)” ^[4]. In other words, a single question can retrieve and join data from Salesforce plus other tools (marketing platforms, databases, etc.) in one response, minimizing round-trips. SyncHub also uses an internal SQL-generation engine to return only the necessary fields, which saves tokens and improves performance ^[4]. According to SyncHub, their connector exposes a comprehensive Salesforce data model (37 endpoints) and handles authentication seamlessly. Once set up, AI agents can answer queries like “How has our customer base changed this quarter?” or “Add 20 new customers to Salesforce”, with SyncHub handling the API calls under the hood.

4. Salesforce Hosted MCP Servers (Pilot)

In mid-2025 Salesforce released its own managed MCP Servers (via AgentExchange) as a pilot offering ^[1]. These hosted servers expose Salesforce APIs directly to AI agents over MCP. For example, Salesforce noted that AI assistants can now potentially “read/write access to CRM records (Accounts, Contacts, Cases, etc.)” through MCP tools (while respecting org permissions) ^[5]. In practice, this means Salesforce organizations can enable native MCP servers (no third-party setup) so AI tools like an approved Claude agent can query or update Salesforce data in real time. This official solution is still in pilot and focused on core CRM clouds, but it promises the most seamless, secure integration by leveraging Salesforce’s own authentication and auditing. In short, Salesforce’s MCP servers turn the org itself into an AI-friendly data source without custom code ^[1].

5. CData – MCP Server for Salesforce

CData, known for its connectors, also offers MCP servers that sit on top of Salesforce. For example, the CData MCP Server for Salesforce exposes the platform’s data via a relational model that LLMs can query in natural language. Their read-only connector “leverages the CData JDBC Driver for Salesforce, exposing Salesforce data as relational SQL models and enabling natural language queries” ^[6]. In practice, this means the MCP server presents Salesforce objects (Accounts, Contacts, Leads, etc.) as virtual SQL tables. An AI like Claude can then run queries (via the MCP JSON-RPC interface) to retrieve exactly the records needed. CData’s approach simplifies complex data by letting the AI speak SQL under the hood, so queries like “total revenue per region” or “recent high-value opportunities” just work naturally. (CData also has beta write-capable MCP servers for full CRUD, but their release notes emphasize the SQL-based read interface ^[6].)

6. AiondaDotCom – MCP-Salesforce

The mcp-salesforce project by AiondaDotCom is a popular open-source MCP server for Salesforce, particularly aimed at Claude users. It is described as a “Complete MCP server for Salesforce integration with Claude Desktop” ^[7]. This connector provides “universal Salesforce integration – works with any Salesforce org, including custom objects and fields” ^[8], and offers “full CRUD operations – query, create, update, and delete any Salesforce records” ^[9]. In other words, once you configure OAuth, the AI can perform essentially any data operation on your org. Aionda’s server automates authentication (the AI agent triggers a browser OAuth flow), learns the org schema, and then handles commands. Key features include dynamic schema discovery (so the AI knows your custom fields) and secure token storage for production use ^{[8] [9]}. In use, an agent could ask “Change John Doe’s title to VP” or “List all contacts in California” and Aionda’s MCP server turns that into Salesforce API calls.

7. Suraj Adsul (UBOS) – Salesforce MCP Server

Suraj Adsul (founder of UBOS) has released an open-source “Salesforce MCP Server” (NPM package @surajadsul02/mcp-server-salesforce). This MCP server similarly connects Claude (or any MCP client) with Salesforce. It advertises a rich feature set: natural-language object/field creation and modification, smart partial-name searching of objects, detailed schema metadata, and flexible record operations. As one description notes, it “supports creating and modifying objects and fields, smart object search, detailed schema insights, flexible record queries, and data manipulation operations” (insert, update, delete, upsert) (Source: www.mcp.pizza). In practice, it runs as a local service (often via an npx command) and handles both basic CRUD and advanced features like SOSL cross-object searches. This makes it a powerful community connector – effectively a complete conversational interface to Salesforce, all installable via NPM (Source: www.mcp.pizza).

8. smn2gnt – MCP-Salesforce Connector

The MCP-Salesforce project by user smn2gnt is another open-source connector. According to its GitHub, it is “an MCP server implementation for Salesforce integration, allowing LLMs to interact with Salesforce data through SOQL queries and SOSL searches” ^[10]. It is a Python-based server that supports executing arbitrary SOQL and SOSL, as well as creating, updating, and deleting records. While less widely publicized than the above products, it provides essential tools like running queries, managing records, and even integrating note-taking. In short, smn2gnt’s MCP-Salesforce offers a lightweight way to give an AI agent full-scope access to Salesforce data and metadata via standard protocols ^[10].

9. MuleSoft Anypoint – MCP Connector

Though not Salesforce-specific, MuleSoft’s Anypoint MCP Connector is worth noting for Salesforce shops. MuleSoft describes this connector as allowing Mule applications to act as both MCP servers and clients ^[11]. In practice, a Mule flow can invoke an MCP server (like Salesforce’s or others) or expose Mule data as tools for an AI. For Salesforce, one could use Mule’s MCP connector to bridge between Agentforce and Salesforce (since MuleSoft is Salesforce-owned) or to integrate an external AI agent into a Muleflow that then updates Salesforce. The key point is that MuleSoft makes any system (including Salesforce via its APIs) available over MCP. The documentation states: “When acting as a client, MuleSoft facilitates the creation of integrations and orchestrations where AI agents are integrated as just another system in the workflow” ^[11]. In essence, the MuleSoft MCP connector is a premium way to orchestrate AI-driven workflows across enterprise systems including Salesforce.

10. syafiq555 – Salesforce MCP Server

The GitHub project salesforce-mcp-server by user syafiq555 implements another MCP interface for Salesforce. This one is relatively simple: it’s described as “a Claude Model Context Protocol (MCP) server implementation that integrates with Salesforce, providing seamless access to Salesforce data through a standardized protocol” ^[12]. In its current form it supports basic queries (e.g. listing Accounts) and can be extended to more objects. Like similar tools, it requires a Salesforce-connected app for OAuth and then exposes selected data via MCP. It demonstrates how the community is building MCP bridges: you run the Node.js server locally, and then your AI (e.g. Claude Desktop) can discover and call tools to query Salesforce. Although initially limited in scope, it reinforces that many open MCP connectors exist to allow natural-language Salesforce queries, and projects like this can be expanded to cover full CRUD.

Each of the above connectors bridges your Salesforce org to AI tools via the new MCP standard. They vary in scope – from full-fledged commercial platforms (Cirra, SyncHub, CData) to Salesforce’s own hosted servers and open-source projects – but all share the goal of letting you “chat with Salesforce”. When evaluating them, consider factors like supported features (read-only vs. full CRUD), ease of setup (hosted vs. DIY), and security (OAuth and permission handling). With any of these MCP connectors, Salesforce data becomes directly summonable by your AI assistant, unlocking faster insights and automation while preserving Salesforce’s security and data governance ^{[1] [2]}.

Sources: Official announcements and product docs ^{[2] [3] [4] [1] [6] [8] [9]} (Source: www.mcp.pizza) ^{[10] [11] [12]}.