---
title: "Exchange: Getting Started with Cint MCP in Claude"
slug: "getting-started-with-cint-mcp-in-claude"
updated: 2026-10-29T19:06:33Z
published: 2026-10-29T19:06:33Z
canonical: "help.cint.com/getting-started-with-cint-mcp-in-claude"
---

> ## Documentation Index
> Fetch the complete documentation index at: https://help.cint.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Exchange: Getting Started with Cint MCP in Claude

> [!CAUTION]
> This feature is currently in **closed-beta**. If you are interested in participating in the beta program, please contact your Customer Success Manager (CSM).

This guide covers: how to ask in plain language, what Claude's reply means, how to compare markets in one request, and how to set up a draft target group without launching it.

## Terminology check

Before getting started, please note that Claude will say "target group" in places where you would say "survey". The distinction is important because it makes Claude’s replies easier to read.

| Term | Definition |
| --- | --- |
| Survey | The questionnaire the respondent answers. It lives on your own survey platform, not on the Exchange. The Exchange only stores its live and test links. |
| Target Group | The Exchange entity that wraps your survey and controls who gets sent to it. It defines the market, qualifications, quotas, completes goal, price, supply, and field dates. |

One survey can carry several target groups. For example, a five-market study is a single questionnaire paired with five target groups — one per market—all under one project.

You don't need to change how you talk about this. You can keep saying "survey," and Claude will follow your lead. The reason to know the difference is so Claude's replies make sense: if it says "three of your five target groups are live," that's telling you three specific markets are live. Not describing your questionnaire itself.

## Run your first feasibility check with Cint MCP

> [!NOTE]
> Feasibility checks and drafts don't purchase anything from Cint. Your LLM token usage still applies.

Once Cint MCP is [connected in Claude](/v1/docs/cint-mcp-installation), start with a feasibility check. It's read-only, it costs nothing (outside your LLM token usage costs), and it's the most common task you will run.

1. **Ask in plain language**

Start a new chat and describe what you need, using your own numbers. **For example:** *Is 500 completes of US parents of under-5s feasible at 12 minutes and 25% incidence? What's the CPI?*

You don't need to name a tool, learn any syntax, or know what an operation is called.
2. **Read the line where Claude says what it sized**

With the result, Claude states what it sized: the market, the number of completes, the interview length, and the incidence rate.

For the audience, Claude keeps your wording and names the attribute that each qualification measures. If the Exchange has no exact match for what you described, Claude marks the attribute as the closest available measure.

**For example:** Sized: US English, 500 completes, 12-minute LOI, 25% IR. “Parents of under-5s” was sized as parents of children aged 0–5.

Read that line. It is the cheapest moment in the whole process to catch a misread—far cheaper than catching it in a launched survey. If anything is wrong, say so, and Claude will size it again.
3. **Look at what comes back**

You will get a verdict and a suggested price. Two things to notice:
  1. **The verdict is about your goal, not an abstract score.** There are three possible outcomes: the whole sample is achievable; only part of it is achievable, in which case Claude tells you how many completes the market can deliver and suggests a way forward, such as a shorter interview, a smaller sample, a wider audience, or a longer field window; or none of it is achievable.
  2. **A price that lands on a round number**—the very bottom or top of the range—is usually the Exchange’s price control rather than a bad estimate. The floor protects suppliers from underpayment; the ceiling protects buyers from runaway costs.
4. **Compare markets in one go**

Try: *Same audience across the US, UK, and Japan—compare them.*

This is where the Exchange is most useful at the proposal stage: several markets, one request, in less time than opening the UI.
5. **Go one step further, still for free**

Try: *Set up a draft target group for the US version, but don't launch it.*

A draft costs nothing (outside your LLM token usage costs). It *is the safe way to see exactly* what would be created. You can also ask for a dry run at any point—example: “dry run that first” —and Claude will show you the exact request it would send, without sending it.
6. **Stop here.**

Launching is a real purchase of sample and cannot be undone. See [Exchange: Cint MCP capabilities, permissions, and troubleshooting](/v1/docs/getting-started-with-cint-mcp-in-claude#launch-a-survey) for more information on launching a purchase sample.

**Example: 3 market feasibility analysis screenshot**

![](https://cdn.us.document360.io/8b64b644-adbd-4719-bdfd-01e2fd9ac6e4/Images/Documentation/image(63).png?sv=2026-02-06&spr=https&st=2026-10-30T10%3A39%3A45Z&se=2026-10-30T11%3A39%3A45Z&sr=c&sp=r&sig=uWfN7M%2FNCZs8To3GrvpgfNXrGOXSAghOZ4eRdwb%2BNsI%3D)

## How to use Cint MCP

> [!CAUTION]
> This feature is currently in **closed-beta**. If you are interested in participating in the beta program, please contact your Customer Success Manager (CSM).

This guide covers common tasks once Cint MCP is connected in Claude: phrasing requests, adjusting how a survey fills, launching a target group, and diagnosing one that isn't filling.

> [!NOTE]
> If you haven't run a feasibility check yet, start with [Getting Started with Cint MCP in Claude](/v1/docs/getting-started-with-cint-mcp-in-claude) before using this guide.

### Ask Claude for what you want

When working within Claude, there is no need to name a specific tool or follow a particular syntax. You also do not need to type a command to start. In any chat where the cint-mcp connector is on, ask for what you need. If you have several connectors switched on, mentioning Cint or the Exchange helps Claude pick the right tools. You can phrase your request however feels natural.

For example:

- *Feasibility for 1,000 completes in Germany, 15 minutes, 30% incidence.*
- *Same audience across the US, UK, and Japan — compare them.*
- *Set up a target group for this brief: [paste the brief]"*
- *How's project 01K3... doing?*
- *Nothing's coming in on the France survey. What's wrong?*

**Give Claude the numbers you actually have:** Give Claude real numbers wherever you have them. Every feasibility check needs the market, the interview length, and the incidence rate. If you leave any of them out, Claude asks for it rather than assuming a value. Other settings, such as field dates, have default values, and Claude reports the values it actually used.

### Incidence rate

State the incidence rate as a plain percentage. **For example:** *"incidence is about 25%."*

Claude will *state the figure it used in its answer*. You can state any incidence rate, including very low ones (such as 1% or lower).

Being able to state a 1% incidence rate isn't the same as it being a sensible buy. At that level, 99 out of 100 people are screened out, and cost, field time, and sample quality all suffer as a result. If your incidence is at or near 1%, treat the feasibility result as a signal to revisit the specification—usually by loosening a qualification or splitting the study—rather than as a price to accept.

> [!NOTE]
> **Always state the incidence rate.** If you leave it out, Claude stops and asks for it rather than guessing. Out of every number in a feasibility check, incidence rate changes the answer the most. A study reported as comfortably feasible at an assumed 50% can be unbuyable at your real 12%. If you genuinely don’t know your incidence rate, say so, and ask Claude to show the result at two or three plausible rates instead.

### Quotas on multi-value targeting

If you request targeting with several values but don't say how to divide them (for example, *"men and women"* or *"18-34 and 35-54"*), Claude will propose a quota split before creating anything. You can respond in one of three ways:

- Accept the proposed split.
- Change the percentages.
- Tell Claude the values are only a filter and shouldn't have quotas.

### Supply, if you need to be specific

You can specify where a target group draws its sample from when you create it: Open exchange, private exchange, a share reserved for a named supplier, or a supplier you want blocked. Claude can list your account's supply allocations, so *"show me our allocations and use the one we normally use for healthcare"* is a reasonable request.

If you don't specify supply, the target group draws on your account's default allocation. For most studies, this is the right choice, so you only need to specify supply when you want something different, such as a reserved supplier or a blocked one.

> [!NOTE]
> Supply can still be changed while a target group is a draft. Once it's in field, changing supply is an Exchange UI action, not something Claude can do.

## Set how a survey fills

You can ask Claude to set pacing on a draft target group or on one that is already live. Describe what you want in plain language. For example:

| What you say | What Claude does |
| --- | --- |
| *"Pace this evenly over the two weeks."* | Releases the completes goal steadily across the field window. |
| *"It's filling lopsidedly, balance the quotas."* | Adjusts pacing so that quota groups fill at similar rates. |
| *"Soft launch 10% first."* | Collects 10% of the completes goal, then pauses so you can review before it continues. |

Before making a change, Claude reads the target group's current pacing settings. It changes only what you ask for.

## Launch a survey

Nothing goes into field until you launch it. Claude creates target groups as drafts, and launching is a separate step that always requires your approval.

### Before you launch

1. Ask Claude to read the target group back to you. It shows the market, qualifications and quotas, completes goal, price, supply allocation, and field dates.
2. Check the price and the completes goal. Those are the two settings that cost money if they are wrong.

### Launch

1. Tell Claude which target group to launch, for example: *"Launch the Germany target group."* Replace "Germany" with the market or name of your target group.
2. Claude asks you to approve the action. After you approve, Claude confirms what went live.

Launching is a real purchase of sample and cannot be undone. Review each approval prompt before you accept it, and do not set Claude to approve launches automatically.

### Launching several markets

A multi-market study has one survey and one target group per market, all under one project. You can launch all of them at once (*"launch all five target groups"*) or one at a time, so you can check how the first one fills before launching the others.

Launching one target group does not launch the others. A target group left as a draft stays a draft—and costs nothing outside your LLM token usage costs—and can be launched later.

### After launch

Claude can pause, resume, and close a live target group, and it can change the CPI and pacing.

Changing field dates or supply on a live target group must be done in the Exchange UI.

### If a launch fails

A launch usually fails for one of three reasons:

- The target group has no live URL.
- The price is zero on a dynamically priced business unit.
- A qualification still needs quotas before the target group is complete.

Claude tells you which one applies, and you can fix it in the same conversation.

## Diagnose a survey that is not filling

If a survey isn't collecting completes, ask Claude what's wrong and name the survey or market. For example, *“Nothing's coming in on the France survey. What's wrong?”*

Claude reads the target group's current state before it suggests anything.

Fixes, in the order Claude works through them:

1. **Check the targeting.** Look for qualifications that are too restrictive.
2. **Extend the field window**, if your deadline allows. This is often the cheapest fix. Changing field dates is an Exchange UI action.
3. **Rebalance pacing or quotas**, if some quota groups are filling at the expense of others.
4. **Raise the price.** This is the last option, and it applies only to dynamically priced surveys. Review the cost calculation before you approve the change, because a CPI increase applies to every complete collected after it.

**If your business unit uses a rate card**

The price on a rate card cannot be changed. Instead, you can add a price boost under Automate Fielding, which pays above the rate card price to compete for supply. Ask Claude about it if low fill is the problem.

## Getting updates

There is nothing to install. Improvements to how Claude works with the Exchange are delivered by the Cint server, so everyone gets them at the same time without downloading anything.

Start a new chat to pick up a change. An open chat keeps the instructions it received when it started.

If you installed an earlier round of the beta, make sure the old pieces are gone: the Cint Exchange MCP connector and the Cint Exchange plugin (see Clear out earlier versions). Leaving either in place gives Claude two conflicting sets of tools or instructions, doubled approval prompts, and results that do not agree with each other.

To check, open your connectors list and confirm that cint-mcp is listed and Cint Exchange MCP is not. Then open your plugins list and confirm that no Cint plugin is installed.

Follow Claude’s current instructions → [Get started with connectors](https://claude.com/docs/connectors/getting-started) and [Use plugins in Claude](https://support.claude.com/en/articles/13837440-use-plugins-in-claude)