Connect AI assistants to Augela (MCP)

Use your company's agents and confirmed knowledge from the AI assistants your people already work in: Claude, ChatGPT, Microsoft Copilot Studio, Claude Code and any other assistant that supports MCP (Model Context Protocol, the open standard these assistants use for connectors).

Updated 25 October 2026 · Available on every plan

What it does

Augela is an MCP server. Once connected, a person can list the company's agents, search the company's knowledge, ask an agent and report a wrong answer, all from inside their assistant and all within their own access in Augela. Corrections made in Augela reach every assistant that asks afterwards.

Server address, the same for every company:

https://interface.augela.com/api/mcp

Before you start

A Company admin turns the connector on in Augela, in Configuration → Connectors → Claude, ChatGPT, Copilot, …, and confirms that answers go to the assistant that asked, under the company's agreement with that vendor.
Admins choose which agents can be asked, per Company, Group or agent (Inherit, enabled or disabled). Group admins can only narrow access, never widen it.
Each person needs an Augela account in that company.

Connect Claude

For Claude on the web, desktop or mobile:

Open Settings → Connectors and choose Add custom connector.
Enter https://interface.augela.com/api/mcp and choose Connect.
Sign in with your Augela account. If you belong to more than one company, choose the company.
Review the consent page. It shows the company, what Claude may do, and that answers are sent to Claude. Choose Allow.

No key is needed. Your Augela password is never shared with Claude.

Screenshot: the consent page, "Claude wants to use Augela as you".

Connect ChatGPT

Where your ChatGPT plan allows custom connectors, add a connector with the same address, https://interface.augela.com/api/mcp, then sign in with Augela and allow access, as for Claude.

Copilot Studio, Claude Code and scripts

In Augela, create a key for the connection.
In Microsoft Copilot Studio, Claude Desktop, Claude Code or your script, add an MCP server with the address https://interface.augela.com/api/mcp and the key.

The key decides the company and the person it acts as. Keep it secret, and revoke it in Augela at any time.

The four tools

List the company's agents: Shows the agents this person may ask, with what each is for. Assistants call it first when they do not know which agent to ask.
Search the company's knowledge: Returns confirmed facts first, labelled "Confirmed by …" or "not rechecked since …", then document excerpts with their sources. Uses no AI on Augela's side.
Ask one of the company's agents: Returns the agent's full governed answer with its sources, as in Augela itself. Follow-up questions continue the same conversation.
Report a wrong answer: Sends the answer to the company's Review Queue, like a thumbs-down in Augela's chat. A reviewer corrects it once, and every assistant gets the corrected fact. The reporter sees the outcome in "What came of my reports".

Screenshot: an answer in Claude from Augela, with "Confirmed by …".

Governance and limits

Off by default, until a Company admin turns it on.
Each person sees only their own access. Nobody sees another Group's memory.
Every question is checked against the company's watch words before anything is searched.
Answers follow the company's governance mode. In Proactive mode, a flagged answer is held for review and never sent to the assistant.
Limits per person and per company. Answers count toward the company's AI budget and spending cap.
Full audit trail: who connected which assistant, every call, every refusal and every report.
Answers given to assistants appear as their own channel, "AI assistants", in the Governance Health Report.

Managing connections

Each person sees their own connections at interface.augela.com/connect and can disconnect any of them at any time.
Admins see who is signed in from which assistant, can sign anyone out, and can revoke keys instantly.
In Configuration → Connectors, the connector shows "On · N agents enabled · N signed in" while active.

Data handling

What Augela receives: the question or report the assistant sends, and the identity of the signed-in person or key.
What Augela sends back: the answer, confirmed facts and document excerpts with their sources, within that person's access. Memory titles and who confirmed them are included for signed-in people.
Where answers go: to the assistant that asked. Its vendor handles them under its own terms and your company's agreement with that vendor.
What is recorded: every call, refusal and report, in your company's audit trail in Augela.

More in the Privacy Policy, the Terms of Service and the Augela Trust Center.

Support

Questions, problems or feedback: human@augela.com. A person answers, usually within one working day. Security reports: security@augela.com.

Also see Augela in your AI assistants and MCP integration, which covers the other direction: Augela agents calling your own MCP servers for live data.