A managed MCP server that lets AI tools like Claude, Cursor, and custom agents query and operate on your Salesforce org via the standard Model Context Protocol. Multi-tenant, secure, and ready to deploy.
A secure, multi-tenant MCP server that exposes Salesforce data and operations to any AI agent speaking the Model Context Protocol.
SOQL queries, SOSL search, object describe, tooling API, composite requests, bulk queries, and metadata deploy — all exposed as standard MCP tools.
Each tenant gets their own API key, Salesforce credentials, tool allowlist, and quota limits. Full data isolation with shared infrastructure efficiency.
Four Salesforce auth strategies: forwarded token, refresh token, client credentials, and JWT bearer. Secrets resolved from env vars or GCP Secret Manager.
Per-tenant daily and per-minute rate limits. Backends for in-memory, Redis, and Firestore ensure quotas scale from local dev to production.
Read-only and read-write modes per tenant. The free tier is restricted to safe read operations, while higher plans unlock write and deploy capabilities.
Run locally with the MCP SDK stdio server for Claude Desktop and Cursor, or bring your own client. Works over both HTTP and stdio transports.
Three steps to connect your AI agent to Salesforce via MCP.
Call /v1/public/register with your org name and plan. You get back a tenant ID and API key instantly.
Set your Salesforce auth strategy — forwarded token, refresh token, client credentials, or JWT bearer — via the admin panel.
Point your MCP-compatible AI agent (Claude, Cursor, or custom) at /v1/managed/mcp with your API key. Start querying Salesforce.
Register a tenant and start querying your Salesforce org in minutes.