qlaris connector for Claude

The qlaris connector lets Claude run synthetic market research for you: surveys, user tests, interview studies and focus groups on digital-twin panels, and questions to your workspace's research repository. Results come back into the conversation, with a link to each study in the qlaris app.

Last updated: October 2026

What the connector does

qlaris runs research on digital twins: synthetic respondents calibrated to demographic data for a country panel. They are AI personas, not real people, and every finding is a simulation. When you ask Claude to run a study, Claude calls qlaris tools to set it up, launch it and read the results. The study runs on qlaris's servers and is stored in your qlaris account, where you can also open it at app.qlaris.ai.

The connector also reads the company context stored in your qlaris workspaces and, where question answering is enabled for your account, the research documents stored there, so Claude can answer questions from your own research.

Connect qlaris to Claude

From the Claude connector directory

Open Claude's connector settings, find qlaris in the connector directory and choose Connect. Claude then opens the qlaris sign-in page (see Sign in). On Team and Enterprise plans, an owner may need to allow the connector for the organization first.

As a custom connector

In Claude's connector settings, choose Add custom connector, name it qlaris and enter this URL:

https://app.qlaris.ai/api/mcp/claude

Then choose Connect and sign in.

In Claude Code

claude mcp add --transport http qlaris https://app.qlaris.ai/api/mcp/claude

Then run /mcp in Claude Code and choose qlaris to sign in in your browser.

If you connected qlaris earlier with https://app.qlaris.ai/api/mcp (the address on qlaris.ai/mcp), that connection keeps working. This page describes the connector at https://app.qlaris.ai/api/mcp/claude, the one listed in the Claude connector directory.

Sign in

The connector uses your qlaris account and signs in with OAuth. Claude opens a qlaris page where you sign in with your email address and password or with Google, then approve access. The approval page names the application that asks for access. Claude never sees your qlaris password.

If you do not have a qlaris account yet, create one at app.qlaris.ai first, with the same email address or Google account you will use to sign in from Claude.

Validation units

Launching a study spends validation units from your qlaris account, or from your organization's shared pool if you belong to a qlaris organization. Listing, reading results, drafting an interview study, dry runs and the research validity check spend nothing. Before a launch, a dry run returns the cost of a survey or a user test, and the draft returns the price of an interview study.

Nothing can be bought in Claude. Units are added in the qlaris app. If your balance is too low, the launch is refused and nothing is charged.

Tools

The connector offers the tools below, grouped by what they do. Which tools you see depends on the features enabled for your qlaris account and organization, so some groups may not appear for you.

Help

Lists what qlaris can do and which tool fits each task.

help_with_qlaris

Audiences and personas

Shows the available country panels, audience filters and segment panels, and manages personas: reusable audience definitions written in plain language. Archiving a persona is reversible.

list_countries list_filters list_segments create_persona list_personas get_persona archive_persona restore_persona

Workspaces and context

Creates workspaces to group your research, and reads and edits the company context stored for a workspace (strategy, positioning, goals and similar entries). Archiving an entry removes it from the active context.

create_workspace list_workspace_contexts get_workspace_context list_context_entries create_context_entry update_context_entry archive_context_entry

Surveys

Runs a survey of up to 10 questions on a panel, then reports its progress, its results with statistics and an AI-written analysis, and its definition. A dry run returns the cost and starts nothing.

list_question_types run_survey get_survey_status get_survey_results get_survey list_surveys

User tests and prototypes

Launches a user test on a public URL: AI participants try a task in a real browser and think aloud, and the results report where they succeeded and where they got stuck. Prototypes are HTML pages qlaris hosts at a public URL, for example to run a user test on them; their files arrive through upload links.

run_user_test get_user_test_status get_user_test_results list_user_tests cancel_user_test request_asset_upload serve_prototype update_prototype_assets list_prototypes delete_prototype

Interview studies

One-to-one interviews with many participants from one discussion guide. The draft is free and returns a complete study with its price; launching it spends units. After completion you can ask the study one more question, follow up with one participant, regenerate the report, or cancel a running study.

draft_interview_study run_interview_study get_interview_study ask_interview_study manage_interview_study list_interview_studies

Focus groups

One shared discussion between three to eight participants, moderated or unmoderated, built turn by turn, with a summary when it ends.

start_focus_group advance_focus_group get_focus_group list_focus_groups

Research repository

PDF research documents can be added to a workspace, checked while they are processed, and removed. Where question answering is enabled for your account, query_knowledge answers questions from those documents, with every statement tied to a numbered source; without it, query_knowledge is not listed.

query_knowledge ingest_research_document get_ingest_status remove_research_document

Research validity check

A free check of how far a research question, or a finished study, can be trusted with synthetic respondents. It returns HIGH, DIRECTIONAL or LOW with the reasons behind the label. You can also tell the qlaris research team that a verdict is wrong.

check_research_validity report_verdict_disagreement

Real human research request

Sends a request to the qlaris team to run a question with real human respondents. The request carries the validity verdict, its reasons and a short research brief. The tool quotes no price and charges nothing; a person from the qlaris team follows up with you.

request_real_research

Decision log and design system

Where these features are enabled for your account: a log of product decisions per workspace, and a stored design system (colours, typography, components, reference screenshots) for building UI. Design-system pages and images are added by URL.

create_decision list_decisions update_decision delete_decision get_design_system add_design_system_guidance remove_design_system_content

Confirmations and side effects

Tools that only read are marked read-only, and Claude can run them without asking. Every other tool is marked as a write tool, so Claude asks you to confirm before it runs, unless you have chosen to allow that tool always. This includes tools that only add something, such as creating a persona, a workspace or a context entry.

Some write tools have a preview inside them. Because the preview is part of a tool that can write, Claude asks for confirmation for the preview too:

run_survey with dryRun: true (returns what would happen and changes nothing)
run_user_test with dryRun: true (returns what would happen and changes nothing)
ask_interview_study with dryRun: true (returns what would happen and changes nothing)
remove_research_document without confirm (returns what would happen and changes nothing)
remove_design_system_content without confirm (returns what would happen and changes nothing)

Three read-only tools keep a small record or have a side effect inside qlaris:

check_research_validity, when it checks research questions before a study, keeps a record of the verdict in qlaris's verdict log (checks of a finished study are logged there too).
check_research_validity, when it checks a finished study, records the verdict with that study.
query_knowledge, where it is listed for you, saves a receipt of each answer in your workspace for 30 days (ids and counts of what the answer showed, with your question stored only as a keyed hash) and, in a research repository whose documents are tagged by subject, a support log record of how the answer's documents were screened, which includes your question.
get_focus_group, when it reads a focus group that has ended, can send that focus group's completion email to the person who started it, if the email has not been sent yet.

The "with qlaris" phrase and the qlaris Claude plugin

Start a message with with qlaris: (for example, "with qlaris: test our sign-up page with first-time visitors in Germany") and Claude uses help_with_qlaris to see what qlaris can do and which tool fits your request. Where query_knowledge is listed for you, start a message with qlaris knowledge: to ask your workspace's research repository a question.

The qlaris Claude plugin is optional. It bundles the same connector with the help-with-qlaris skill, a fuller guide Claude can follow for qlaris tasks. In Claude Code, install it from the qlarisai marketplace:

/plugin marketplace add qlarisai/qlarisai-claude-plugin
/plugin install qlaris@qlarisai

If you already use the qlaris plugin in Claude Code, its update moves it to this connector (/api/mcp/claude) and adds the skill. After the update, Claude Code asks you to sign in to qlaris once more: run /mcp and sign in with your qlaris account. The plugin's README describes installing and updating in more detail.

Data and privacy

When Claude calls a qlaris tool, it sends that tool's arguments to app.qlaris.ai: for example your research questions, study content, audience criteria, the URLs to test and the files you upload. qlaris stores them in your qlaris account, together with the studies they create. qlaris receives what Claude sends in tool calls, not your whole Claude conversation.

Responses are generated by AI models; qlaris uses third-party AI providers to run them.
A user test visits the URL you give it in automated browser sessions.
qlaris records which tools are called and when, for usage statistics and support.
To disconnect, remove the connector in Claude's connector settings. You can also end connected sessions in the qlaris app settings.

The privacy policy describes how Qlaris GmbH processes personal data and how to exercise your rights. The terms of service apply to the connector as to the rest of the qlaris platform.

Support

For help with the connector, questions about your account or a problem with a study, contact the qlaris team through the contact page. Privacy requests go to data@qlaris.ai and security reports to security@qlaris.ai.

Troubleshooting

Reconnect

If qlaris tools stop answering, or Claude says a tool is not available, disconnect qlaris in Claude's connector settings and connect it again, then start a new chat. This reloads the tool list and renews the sign-in.

Sign-in errors

Sign in with the same method you use in the qlaris app: email and password, or Google. If the approval page shows a different qlaris account than you expected, sign out of qlaris at app.qlaris.ai and connect again.
Allow pop-ups from Claude if the sign-in window does not open.
If the approval page shows an error or expires, close it and choose Connect again in Claude to start a fresh sign-in.

A study is refused

A launch can be refused when your balance is too low, when the audience matches too few panel members, or when a study type is turned off for your organization. The tool's answer says which, and nothing is charged.

File uploads fail in claude.ai

Prototype files are uploaded from Claude's code-execution sandbox, which reaches only allowed hosts. If an upload fails with a network error, add app.qlaris.ai under Settings, Capabilities, Code execution and file creation, Additional allowed domains.