MCP Server

Every enneo instance is also an MCP server. You connect your AI client — Claude, ChatGPT, or any other client that supports the Model Context Protocol — using a single address, and the assistant then works with the API of that very instance: researching tickets, checking configuration, pulling evaluations and, if you allow it, making changes.

There is nothing to install — no extension, no marketplace, no local working copy. You enter one address in the client and sign in through the browser with your usual enneo account:

https://<your-instance>.enneo.ai/mcp

Investigate tickets

Search, read and evaluate tickets — status, history, customer recognition, detected tags and intents, processing traces.

Find this instance's API

The assistant searches your own instance's API description, so it knows exactly the endpoints, fields and required values that apply to you.

Guided procedures

For recurring questions — "Why didn't automation fire?", "What is the customer actually asking for?" — the assistant follows a fixed procedure instead of improvising.

Query the documentation

Answers from the published enneo documentation — in the same conversation, without switching windows.

Read-only or full access

This is the decision you make before connecting — and the only one you cannot change later in the client, but only by creating a new connection.

When you confirm in the browser there is an Allow changes checkbox. It is not ticked by default: anyone who simply clicks "Allow" gets a read-only connection, even if the client asked for more.

Capability	Read-only	Full access
Find this instance's API operations	✓	✓
Read API calls (tickets, customers, settings, events …)	✓	✓
Read product documentation	✓	✓
Read procedures	✓	✓
Write API calls (create, change, delete)	—	✓

A read-only connection is not quietly narrowed: an attempt to write is refused, and the assistant is told in the refusal that the connection is read-only.

Warning

The assistant acts as the person who created the connection, and therefore sees exactly what that person is allowed to see in enneo — no more and no less. So connect a client with an account whose permissions match the job.

Info

A ticket the assistant reads is always recorded as not viewed: no time tracking, no change to who worked on the ticket. Research therefore does not distort your team's handling statistics.

Setting up the connection

Enter the address in your AI client

In your client, add a new MCP server (some clients call it a "connector") with the address https://<your-instance>.enneo.ai/mcp. No identifier, key or configuration file is required.

Sign in through the browser

The client opens a browser window with your enneo instance's sign-in page. Sign in as usual — if you are already signed in in that browser, the step is skipped.

Confirm access

A page appears saying "<Client> wants to access enneo", naming the account the client will act as. Tick Allow changes only if the assistant really should be able to write, then confirm with Allow.

Done

The client returns automatically and is connected. The connection carries the name of your instance — "Enneo Acme", for example — so you can reliably tell several instances apart inside the same client.

Reviewing and revoking connections

Your own profile has a Connected assistants section. It lists every connection with its access level, the date it was created and when it was last used — and lets you revoke it. A revocation takes effect immediately: the client's next call fails, and the connection cannot be restored. The client then has to connect again.

Examples

You can copy the following prompts into your client unchanged.

Investigate tickets

Investigate ticket #5678: status, history, recognized customer, tags and intents —
and finish with two sentences on where it is stuck.

Show me the open tickets from the last 24 hours on the email channel
and group them by detected request.

What did the customer in ticket #1234 actually ask us to do, and are they
sufficiently legitimated for us to carry it out?

Understand automation

Why did no AI agent fire on ticket #5678?

Why wasn't the tag "contract termination" detected on ticket #4321?

Trace the processing of ticket #5678 — which phases ran, what came out of
them, and were there any alerts?

Check configuration

Which email subchannels are configured, and which one receives
support requests?

Show me the settings that control auto-processing and explain their
current values.

Which user-defined functions are also available as AI tools?

Telephony

How was the call in ticket #9001 routed, and why did it end up with the
voicebot instead of an agent?

Documentation and background

How is customer recognition by contract number configured?
Answer from the enneo documentation.

Which conditions have to be met for a rule-based agent to run automatically?

Procedures as slash commands

enneo ships ready-made procedures for the most common investigations — among them: investigating a ticket, working out why automation did not fire, tracing the processing of an event, checking tag detection, analyzing a call, and finding out what a customer asked for.

Clients that support MCP prompts show these as slash commands — in Claude Code, for example, as /mcp__<server>__<name>. You pick the command, supply the ticket number, and the assistant works through the procedure step by step. In every other client, the assistant reaches for the same procedures on its own.