MCP App Store
by KeenEthics
Persona icon

Persona MCP Server for ChatGPT

By Convosight

Overview

Convosight's Persona gives you simulated consumer personas built from real social and consumer research. Browse the personas and ask them questions to get realistic opinions and reactions to ads, messaging, products and ideas. Pick a persona, ask your question, and read the answer in an interactive chat widget. Open the full report for any conversation. Limitations: answers are AI-simulated persona responses, not real survey data, and a daily question limit applies per user. Requires signing in with a Convosight account.

How to connect

  1. ChatGPT: Open the Persona listing in the ChatGPT app directory and choose Connect. ChatGPT listing Persona exposes 7 tools here.

Tools

ask_persona

ChatGPT
Sends a message to one of this team's Convosight personas — a simulated consumer built from real social/consumer research — and gets its in-character reply. Call list_personas first if you don't already know the persona_id — and if that returns more than one candidate, do NOT guess and call this yourself; let the user pick via the list_personas widget, unless they then name one exact persona themselves. A pick made in that widget reaches you as widget state / model context labelled "Persona widget state" with persona_id and session_id (or selected_persona_id when no message was pending) — use exactly those values here. RULE — STICK WITH THE ESTABLISHED PERSONA: once ask_persona has been called for a given persona_id in this conversation, or the user picked one in the list_personas widget (see above), that persona is the default for everything that follows, including short or vague follow-ups and new topics — call ask_persona with the SAME persona_id and session_id rather than list_personas, unless the user explicitly asks to talk to a different persona or switch. EVERY FOLLOW-UP GETS ITS OWN CALL: you never receive the persona's actual reply text back from ask_persona — only {jobId, session_id} — so you have no real reply of your own to answer a follow-up from. Never answer a message directed at the persona yourself; always call ask_persona again for it, even a short or vague one. A message that builds on an earlier ask_persona call to the SAME persona_id shares context with it via the session_id from that earlier call — see the session_id parameter for details. Returns {jobId, session_id} for a reply generated asynchronously — this response itself carries no reply content. Where this tool has an associated widget, the widget displays the full reply as it streams in. A result of quotaExceeded: true means the daily message limit has been reached; alreadyInProgress: true means a message on this session_id is already being processed. Neither reflects a completed reply — the widget already displays the relevant message to the user.

ask_persona_widget

ChatGPT
Internal tool called directly by the persona widget (via its host bridge) when the user picks a persona from the list_personas picker. Starts the persona chat and returns { jobId, session_id } so the widget can stream the reply in-place. Not intended to be called by the model.

get_persona_answer

ChatGPT
Polls the status of a job started by ask_persona or ask_persona_claude. Used internally by the persona widget via the host's own tool-calling bridge, and also the fallback for models on clients that do not render the widget: call this with the jobId and repeat until status is 'done', 'error', or 'not_found' (the latter two are terminal — stop polling and relay the outcome instead of retrying). Blocks server-side up to 25 seconds per call. Returns {status, content?, artifacts?, session_id?, quota?}.

get_persona_chat_history

ChatGPT
Internal tool called directly by the persona widget (via its host bridge) to restore a previously-answered message that isn't cached locally. Not intended to be called by the model.

list_personas

ChatGPT
Lists the personas this team can chat with, rendered as a clickable picker where this tool has an associated widget. Call this first when the user wants to talk to a persona but hasn't given (or you don't already know) its exact persona_id — e.g. they named a persona you can't uniquely resolve, or just said "ask a persona" without saying which. Pass "search" to find a persona by name; omit it to see all available personas (public, team and your own), most recently used first. Pass "question" too if the user already gave one — the widget lets them pick a persona and sends the question to it directly. If "search" and "question" are both given and exactly one persona matches, the chat starts immediately and this returns the same {jobId, session_id, persona_id} shape as ask_persona — treat it exactly like an ask_persona result (the widget streams the reply). The same happens when this conversation already has an established persona (picked in this tool's widget, or asked via ask_persona) and "question" is given without "search": the question goes straight to that persona in its existing session and no picker is shown. Pass switch_persona: true only when the user explicitly asks for a different persona or to pick again — that always shows the picker. IMPORTANT: after this returns, STOP and let the user pick from the widget — do NOT call ask_persona yourself with a guessed persona_id from the results. Only call ask_persona directly, skipping the picker, if the user then names one exact, unambiguous persona themselves. RULE — DON'T RE-PICK AN ALREADY-ESTABLISHED PERSONA: once a persona_id has started chatting in this conversation (via ask_persona, this tool's own search+question auto-start, or a pick in this tool's widget — which reaches you as widget state / model context labelled "Persona widget state" carrying that persona_id and session_id), that persona stays the target for everything that follows — including a new topical question, not just an obvious follow-up — unless the user explicitly asks to talk to a different persona or pick again. Call ask_persona with that established persona_id instead of calling list_personas.

select_persona_widget

ChatGPT
Internal tool called directly by the persona widget (via its host bridge) when the user picks a persona from the list_personas picker with no message pending yet. Not intended to be called by the model.

Capabilities

WritesInteractive

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/persona"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/persona)