Compatible MCP Clients

Connect your favorite AI to 590+ managed MCP servers. Managed authentication, enhanced execution, and 36,000+ secured and governed MCP tools.

Book a Demo Start Free

Find your MCP client

Allow quick IT setup and instant AI tooling rollout.

Amazon Quick

Web AWS's AI assistant for work

Bedrock AgentCore

Cloud platform AWS runtime for deploying agents

ChatGPT

Web ChatGPT in the browser

ChatGPT desktop app

Desktop ChatGPT for macOS and Windows

Claude (web)

Web Claude in the browser

Claude Code

CLI Anthropic's agentic coding tool

Claude Cowork

Desktop Delegate multi-step work to Claude

Claude Desktop

Desktop Claude for macOS and Windows

Codex

CLI OpenAI's agentic coding tool

Cursor

IDE AI code editor with a built-in agent

Databricks

Cloud platform Genie agents over lakehouse data

Dust

Web Build agents over company data

Fin

Embedded Intercom's AI support agent

Gemini CLI

CLI Google's terminal AI agent

Gemini Enterprise

Web Google's enterprise AI assistant

Gemini Enterprise Agent Platform

Cloud platform Google's agent build and deploy platform

Glean

Web Enterprise search and assistant

goose

CLI Open-source local AI agent

LangSmith Fleet

Web No-code agent workspace for teams

Microsoft 365 Copilot

Web AI assistant across Microsoft 365

Microsoft Copilot

Web Microsoft's consumer AI assistant

Microsoft Foundry

Cloud platform Microsoft's platform for building agents

n8n

Cloud platform Workflow automation with AI agents

North

Web Cohere's private enterprise AI workspace

Notion AI

Embedded AI assistant inside Notion

Perplexity

Desktop AI answer engine and browser

Postman

Desktop API platform that calls MCP servers

Retool Agents

Cloud platform Build business-process agents on Retool

Rovo

Embedded AI agents across Jira and Confluence

SAP Joule

Embedded SAP's business AI assistant

ServiceNow AI Agents

Embedded AI agents on the ServiceNow platform

Snowflake CoWork

Cloud platform Conversational agent over Snowflake data

Visual Studio Code

IDE Microsoft's editor with agent mode

watsonx Orchestrate

Cloud platform IBM's agent orchestration platform

Zendesk AI agents

Embedded AI agents that resolve support tickets

MCP client FAQ

An MCP client is the AI application someone works in, such as Claude, ChatGPT, Cursor or Glean. The Model Context Protocol is how it connects to MCP servers and calls the tools they expose. The client holds the conversation. The server holds the tools and the credentials.

MCP clients that work with StackOne include Claude (web), ChatGPT, Gemini Enterprise and Microsoft Copilot in the browser; Claude Desktop, ChatGPT desktop app, Postman and Perplexity on the desktop; Claude Code, Cursor, Visual Studio Code, goose, Codex and Gemini CLI for engineers; Glean, Microsoft 365 Copilot, Amazon Quick, Dust, North and LangSmith Fleet as company-wide workspaces; Rovo, Notion AI, Fin, ServiceNow AI Agents, Zendesk AI agents and SAP Joule built into SaaS; Microsoft Foundry, Gemini Enterprise Agent Platform, Bedrock AgentCore, watsonx Orchestrate, Retool Agents, n8n, Databricks and Snowflake CoWork as cloud platforms.

Employees can authorize their AI with their own app accounts rather than a shared service account. The employee signs in to StackOne through their AI client, links the accounts they choose, and selects which actions the AI is allowed to run. Every call then executes under that employee's own identity and inherits their existing permissions, so the AI cannot reach anything the person could not reach themselves.

One MCP client can connect to multiple MCP servers at once, and most do. The practical limit is context, not the protocol. Every connected server advertises its tools to the model, so a client attached to a dozen servers spends much of its context window on tool definitions before any work begins. Routing through a single server that exposes only the tools the agent needs avoids that.

The difference is which side holds the tools. An MCP client is the application a person works in. An MCP server exposes tools to that client, handles authentication, and runs each call against the underlying API. StackOne is the server, so the client never holds a credential.

An MCP client that is not listed here will still work if it supports the protocol. This list covers the clients StackOne has tested and documented, which is narrower than what connects. StackOne exposes a standard MCP endpoint, so any compliant client can use it.