Certify now has an MCP server, so you can connect your Certify account to Claude, ChatGPT or Microsoft Copilot and work with your credentials in plain English. Ask "has this person received their badge yet?" or "issue the course badge to these twelve people", and the assistant does it in Certify for you, showing you a preview and waiting for your confirmation before it issues anything.
What is an MCP server?
An MCP server is the piece that lets an AI assistant use another product's tools safely. MCP, the Model Context Protocol, is an open standard for connecting AI assistants to the systems where your data lives. Anthropic introduced it in November 2024, and Claude, ChatGPT and Microsoft Copilot all support it today.
Without MCP, an assistant can only talk about Certify in general terms. With Certify's MCP server connected, it can sign in as you, read the credentials your account can see and, if you're allowed to, issue new ones. The server address is the same for every assistant:
https://api.certify.one/mcp
What can the AI do in Certify?
It can look up your account, your credential designs and the credentials you've issued, and it can preview and issue credentials. The server offers six tools. The four look-up tools are available to everyone who connects; the two issuing tools only appear for people whose Certify account is allowed to issue.
| Tool | What it does | Type |
|---|---|---|
View my Certify account (whoami) |
Shows which Certify account and organisation the assistant is connected to | Look up |
List credentials (list_credentials) |
Lists the credentials set up in your account | Look up |
Find issued credentials (find_issued_credentials) |
Searches the credentials you've already issued | Look up |
View an issued credential (get_issued_credential) |
Opens the details of one issued credential | Look up |
Preview issuance (preview_issuance) |
Shows what would be issued, and to whom, before anything happens | Issue |
Issue credentials (issue_credentials) |
Issues the credentials once you've confirmed the preview | Issue |
In practice, that turns everyday admin into a single question. A training coordinator can check whether a learner's badge went out without opening the dashboard. A programme manager can paste a list of names and email addresses from a spreadsheet and have the right badge issued to all of them, the same job that bulk issuing in Certify handles through a CSV upload.
How does Certify stop the AI issuing the wrong credential?
The assistant always shows you a preview and asks for your confirmation before it issues anything. Issuing is a two-step process by design: the assistant first calls the preview tool, shows you which credential will go to which recipients, and only calls the issue tool after you reply to approve it.
Three further safeguards sit around that:
- It signs in as you. You connect with your own Certify email and password, so the assistant works within your account rather than with a shared key.
- You choose what it may do. When you sign in, you decide which actions the assistant is allowed to take on your behalf.
- Issuing is for issuers only. If your account can't issue credentials, the issuing tools don't appear at all, so the assistant can't offer them.
One thing to check before you start: anything the assistant looks up in Certify, such as a recipient's name and email address, is passed to the AI provider to answer your question. Make sure using Claude, ChatGPT or Copilot with learner data fits your organisation's AI and data protection policies (this isn't legal advice).
How do you connect Certify to Claude?
It takes about a minute and needs no code or API token. You add Certify as a custom connector, paste the server address and sign in. Anthropic's guide to custom connectors using remote MCP covers which Claude plans include them.
- In Claude, open Customize, then Connectors, and click Add, then Add custom connector.
- Name it Certify, paste
https://api.certify.one/mcpas the server address, and click Continue, Add and Connect. - Sign in with your Certify email and password, and choose what Claude may do.
Then try a first question, such as "Which Certify account am I connected to?", to check the connection works.
Can you use Certify with ChatGPT and Microsoft Copilot?
Yes. Both support MCP servers, and you use the same Certify server address. The steps differ by assistant, and your Certify dashboard shows them for each one.
- ChatGPT adds MCP servers as custom connectors. Whether you can add one depends on your ChatGPT plan and, in a company workspace, on your administrator; OpenAI's help article on developer mode and MCP connectors sets out the current rules.
- Microsoft Copilot connects to MCP servers through Copilot Studio, where they become tools for an agent. Microsoft's announcement that MCP is generally available in Copilot Studio explains the approach; it is usually set up by IT.
Where does the MCP server fit with Certify's other integrations?
Use the MCP server for questions and one-off jobs, and integrations for anything that should happen automatically every time. An assistant is ideal when a person is in the loop: checking a record, issuing to a list someone has just sent you, or answering a colleague's query. When a credential should be issued whenever a learner completes a course, with nobody involved, use one of Certify's integrations with learning platforms instead.
If your organisation is rolling out AI tools more widely, it's also worth thinking about how you'll evidence the skills people build with them. Our post on recognising AI skills with digital credentials looks at how organisations are doing that.
