Last updated: 22 July 2026
Audience: Participants using Verity for practice and coaching.
Missed the setup step? If you haven’t connected Verity yet, start with Connect MCP before continuing.
Run Verity roleplays directly in your chat assistant, receive your Evals 2.0 debrief, and retry the moment that had the biggest effect on your score.
Practise and improve with MCP
How do I start a roleplay?
Tell Claude what you want to work on, for example "I want to practise the pricing conversation." It lists the bots you are assigned, starts the session, and shows the bot's opening line.
You type your turns and Claude relays the bot's replies.
Is the scoring really Verity's, or Claude's?
It’s always Verity's. The persona is held on Verity's side, so the character cannot be talked out of role, and the evaluation is the same one the product runs. Claude carries the messages back and forth. It does not act the character and it does not mark your work.
What do I get when I end the session?
Verity scores the conversation exactly as it does in the product, and the full Evals 2.0 result comes back into the chat: the headline score, the four-beat debrief, your dimension scores, the signals with the evidence behind them, and the stats. Scoring runs in the background, so give it a moment after you end before the result lands.
Can I practise the moment that decided my score?
Yes. The debrief names the moment that turned the result. Ask to retry it and you are dropped back into the conversation at that exact point, with the bot's line waiting.
Can I do a voice or avatar roleplay in the chat?
Not inside the chat yet. Chat assistants cannot run live voice while they are connected to tools. Ask for a voice or avatar session and Claude hands you a link into the Verity app; do the session there, then come back and ask "how did I do?" and the debrief arrives in the chat.
What if my session gets interrupted?
It resumes. An in-flight conversation survives a reload or a dropped connection and picks up where you left off.
Can anyone else see my roleplays or scores?
As a participant, you only ever see your own sessions and results through the connector. Your admins and account managers can see team performance, including your scores, because reporting is part of their role, the same as in the product.
If your organisation has privacy settings switched on, such as under a works council agreement, your individual results are hidden even from them and only group numbers show.
Quick reference
I want to… | Do this |
Start a roleplay | Describe what you want to practise; Verity lists your available bots |
Get your debrief | End the session and wait briefly for the Evals 2.0 result |
Retry the key moment | Ask to practise the fix and resume at the pivotal line |
Use voice or avatar | Open the Verity app from the supplied link, then return for your result |
Tools available through the connector
Ask in plain English and your assistant chooses the appropriate tool. Every action runs through your own Verity login and permissions.
Tool | What it does |
| Shows the practice bots you are allowed to use. |
| Starts a roleplay — as text in the chat, or as a link that opens the Verity app for a voice or avatar session. |
| Passes your turn to the bot during a text roleplay and brings back its reply. |
| Ends the roleplay and submits it for scoring. |
| Fetches the scored result once it is ready after you end a session. |
| Pulls your most recent practice debrief — the answer to "how did I do?" |
| Drops you back into the single moment the debrief flagged, to try it again. |