About iMario
iMario Skills provides an agent workflow for conducting synthetic audience research through the iMario MCP server. It is for teams testing copy, pages, prices, concepts, or images; it selects a workspace audience, checks study costs before running, and reports responses rather than forecasts.
Key features
- 01Use audiences already in the workspace
- 02Explain study costs and ask before spending
- 03Test copy, pages, prices, concepts, or images
- 04Summarize option counts, themes, and quotes
- 05Report respondent data sources and limitations
Use cases
- 01Test ad creative with a defined audience
- 02Evaluate pricing or a product concept
- 03Run focus groups around selected topics
MCP support
The Skill operates the iMario MCP server. Connect to `https://mcp.imario.ai/mcp` in a compatible app and approve the connection before an agent can conduct audience research.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://mcp.imario.ai/mcpclaude mcp add --transport http imario https://mcp.imario.ai/mcp{"mcpServers":{"imario":{"type":"http","url":"https://mcp.imario.ai/mcp"}}}Open in Cursor{"servers":{"imario":{"type":"http","url":"https://mcp.imario.ai/mcp"}}}Open in VS Code{"mcpServers":{"imario":{"type":"http","url":"https://mcp.imario.ai/mcp"}}}{"mcpServers":{"imario":{"serverUrl":"https://mcp.imario.ai/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project READMEiMario Skills
Agent Skills for iMario. Ask anyone anything before you decide: iMario turns your own customer data and 5.5 billion people in 61 markets into synthetic audiences calibrated on real data, and this skill lets your agent put copy, a page, a price, a concept or an image in front of the audience it is for.
The skill teaches an agent to use the iMario MCP server the way a researcher would: pick the right audience, price the study and ask before spending credits, run it, and report what respondents said with counts and quotes. Never as a forecast.
Connect iMario first
The skill drives the iMario MCP server, so the server has to be connected in your app.
- Server:
https://mcp.imario.ai/mcp - Add it in ChatGPT, Claude, Claude Code, Codex, Cursor or VS Code, then approve the connection in iMario. Setup for each app: Connect the iMario MCP server.
Install the skill
Let the agent do it
Paste this into Claude Code, Codex or Cursor:
Connect iMario to this agent: imario.ai/SKILL.md
The page at that address is a setup sheet, not the skill: the agent registers the server, installs this skill with npx skills add laugherxiao/imario-skills, and stops at the one step only you can do, approving the connection in your browser. ChatGPT and claude.ai do not act on a pasted page; there you add the connector in the app's settings (steps) and upload the skill by hand.
Claude Code
/plugin marketplace add laugherxiao/imario-skills
/plugin install imario@imario
The plugin also registers the server address. Run /mcp, select imario and choose Authenticate to approve the connection.
To install without the marketplace, copy skills/imario into ~/.claude/skills/.
Cursor
This repository is also a Cursor plugin: the skill plus the server address. Clone it into Cursor's local plugin folder (a copy; Cursor skips a symlink to a folder elsewhere), then run Developer: Reload Window:
git clone https://github.com/laugherxiao/imario-skills ~/.cursor/plugins/local/imario
Open Customize to check that the imario skill and MCP server are there, then select the server in Cursor's MCP settings and sign in to approve the connection.
Codex
Copy skills/imario into ~/.agents/skills/, then restart Codex. Codex reads agents/openai.yaml for the skill's name, icon and its dependency on the iMario server; invoke it explicitly with $imario or let Codex pick it from the request.
Claude (claude.ai and the desktop app)
- Zip the
skills/imariofolder. - Open Customize > Skills, select +, then Create skill > Upload a skill, and choose the zip.
Code execution has to be on: Settings > Capabilities on Free, Pro and Max, Organization settings > Skills on Team and Enterprise.
Claude API and the Agent SDK
Skills do not sync between surfaces. To use the skill in your own agent, zip skills/imario and upload it with the Skills API (/v1/skills), then reference its skill_id in the request's container alongside the code execution tool. See Using Agent Skills with the API.
Other agents
The skill follows the open Agent Skills format. Put the skills/imario folder wherever your agent reads skills.
What the skill does
- Uses an audience from your workspace. Never invents one, and never swaps in a different population without asking.
- Prices any study that is not trivial and asks before spending credits.
- Reports what respondents said: counts per option, themes, two or three quotes, and how many answers came from each respondent's own data. Never says a version "will win".
- Runs your own material: an image for a creative test, or a questionnaire or interview guide transcribed into questions.
- Relays limits and errors with the numbers: credits, the workspace's daily limit, audiences it could not use.
- Stays out of desk research, market sizing and data you already have.
Example prompts
- "How would our closed-lost accounts react to weekly onboarding pricing?"
- "Test this banner with German parents of under-fives. What would stop them from clicking?"
- "Run a focus group with UK GPs on these three topics: workload, referral delays, patient messaging."
Layout
skills/imario/SKILL.md: the workflow and the rules that always apply.skills/imario/references/: how to read results, every error code, the user's own material. Loaded only when needed.skills/imario/agents/openai.yaml: Codex metadata (display name, icon, MCP dependency)..claude-plugin/,.mcp.json: the Claude Code plugin manifest and the server it registers..cursor-plugin/,mcp.json,assets/logo.png: the Cursor plugin manifest, the server it registers and its logo.evals/: the scenarios the skill is checked against, including prompts where it must not fire.
Privacy and support
- Privacy policy: imario.ai/privacy
- Terms of service: imario.ai/terms
- Support: [email protected]
The skill itself stores nothing. Studies it starts live in your iMario workspace under the workspace's data terms.
License
Apache-2.0. See LICENSE.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.