Overview
The Reflect MCP connects your Reflect HR data to your favorite AI assistant (Claude, ChatGPT, Copilot, Dust, or any MCP-compatible app). Once connected, you'll be able to query your data in plain language, directly inside your assistant, with no export.
Before you start
To set up the Reflect MCP, you need three things:
An active Reflect account (your usual credentials: email + password, or SSO).
An MCP-compatible AI assistant (Claude, ChatGPT, Dust, Copilot...).
The Reflect MCP server URL:
https://api.app.getreflect.io/mcp
No developer access or API key is required.
Connect the Reflect MCP
The process is the same everywhere: add an MCP server in your assistant, paste the Reflect URL, and sign in with your Reflect account. That's it.
In Claude
Prerequisite: you must be an administrator of your Claude organization to install a custom connector.
Open Organization Settings.
Go to Connectors.
Click Add, then Custom, then Web.
Name the MCP (we recommend calling it "Reflect"), then paste the Reflect MCP server URL:
https://api.app.getreflect.io/mcp.
The MCP is now available to you and your organization. You then just need to install it from your personal connectors.
In ChatGPT
Open Settings, then Connectors (or Apps).
Add an MCP server / custom app.
Paste the URL https://api.app.getreflect.io/mcp.
Sign in with your Reflect account.
In Dust
In Dust, open the configuration page to add a new MCP.
Enter the URL
https://api.app.getreflect.io/mcp.Leave authentication on automatic.
Sign in with your Reflect account.
The MCP is configured and ready to use.
Other tools
Any assistant that lets you add a custom MCP server works: add the URL above and sign in with Reflect.
Your first prompts
Once connected, ask your questions in plain language. When starting out, add "from the Reflect MCP" to your request so the assistant uses the right source. After a few uses, it will remember on its own.
"What HR data can I query?" (great to start)
"What's our headcount by department, and how has it changed over the last 12 months?"
"Show voluntary turnover by team over the year and flag teams above 15%."
"Compare our 2026 payroll budget to actuals by month and by department."
"For each level in our salary grid, show the median salary and the gender pay gap."
Security and permissions
Read-only: the MCP queries your data, it never modifies it.
Permissions respected: each user only sees the data they have access to in Reflect.
Technical detail: authentication relies on OAuth 2.1 with PKCE and supports DCR (Dynamic Client Registration), the same standard used by Notion, Linear or GitHub. Any Reflect customer can connect the MCP to the AI agent of their choice on their own, with no involvement from the Reflect team.
FAQ
Connection failing?
Connection failing?
Check that the URL is exactly https://api.app.getreflect.io/mcp and that you're using your usual Reflect credentials.
The assistant sees no data?
The assistant sees no data?
Your Reflect permissions apply: you only see what you have access to in Reflect.
Which tools are compatible?
Which tools are compatible?
Any assistant that supports custom MCP servers.