Hunch
Calibrated judgments for text: yes/no probabilities, picks from your options, or scores.
Installation
Ajoutez ceci à la configuration MCP de votre client (Claude Desktop : claude_desktop_config.json ; Cursor : .cursor/mcp.json ; VS Code et les autres utilisent la même forme « mcpServers »).
claude mcp add --transport http hunch 'https://hunchsheet.app/mcp' --header 'Authorization: <Bearer token: "Bearer hunch_<40 hex characters>". Get a free key (100 rows) at https://hunchsheet.app.>'{
"mcpServers": {
"hunch": {
"type": "http",
"url": "https://hunchsheet.app/mcp",
"headers": {
"Authorization": "<Bearer token: \"Bearer hunch_<40 hex characters>\". Get a free key (100 rows) at https://hunchsheet.app.>"
}
}
}
}claude mcp add --transport http hunch 'https://hunchsheet.app/mcp/{api_key}'{
"mcpServers": {
"hunch": {
"type": "http",
"url": "https://hunchsheet.app/mcp/{api_key}"
}
}
}claude mcp add --transport http hunch 'https://hunchsheet.app/mcp/try'{
"mcpServers": {
"hunch": {
"type": "http",
"url": "https://hunchsheet.app/mcp/try"
}
}
}Documentation
Tirée du README du projet sur GitHub, actualisée chaque semaine. Les liens ouvrent GitHub.
Sommaire
Hunch MCP
Hunch gives your AI assistant calibrated judgments about text. For each text you send, it returns the probability that the answer to a yes/no question is yes, one option from your own list, a score on your own scale, or several yes/no answers at once. The answers come from Jev, TypeSafe's System One model, which returns numbers instead of generated text, so your assistant can sort, filter and count them.
The server is hosted at https://hunchsheet.app/mcp (Streamable HTTP). This repo holds the install files for Claude Code and Gemini CLI. Setup for Claude, ChatGPT, Cursor and VS Code is at hunchsheet.app/connectors.
Any agent, one line
npx add-mcp https://hunchsheet.app/mcp -yThat adds the server to every agent it finds on your machine (Claude Code, Cursor, VS Code, Codex, Gemini CLI and more). Sign in when the agent asks, or press Start with 100 free rows on the sign-in page: no email, no card. To try Hunch with no account at all, use https://hunchsheet.app/mcp/try instead: ten texts a day are free. Install lines for each client: hunchsheet.app/install. How agents use it: hunchsheet.app/agents.
Tools
| Tool | Returns |
|---|---|
hunch_ask | The probability of yes, 0 to 1, for each text |
hunch_pick | One option from your list for each text, with its confidence |
hunch_score | A position on your levels (0 to n-1), the most likely level, and its confidence |
hunch_multi | One probability per question for each text, up to 10 questions |
hunch_balance | Credits left on your key |
hunch_quote | Credits and dollars a job needs, before you run it. Free, no key needed |
hunch_buy_credits | A checkout link for the person to open. The assistant cannot pay for them |
hunch_get_key | Keyless /mcp/try connection only: a key with 100 free rows, no email |
Good for leads, support tickets, reviews, survey answers, emails and any list of short texts you want judged the same way. It judges text only: no arithmetic, counting or dates. English works best. Put the definition inside the question, for example "Can this person approve a software purchase without asking someone else?" instead of "Is this a decision maker?".
Get a key
The form at hunchsheet.app emails a key with 100 free rows. No inbox? The sign-in page has a free-rows button, and an agent can call hunch_get_key on /mcp/try. After that, 5,000 rows cost $29 and credits never expire. One credit per answered text per question. Blank texts and repeats in the same call are free.
Claude Code
/plugin marketplace add Eddy1919/hunch-mcp
/plugin install hunch@hunch-mcpThen run /mcp, pick hunch and choose Authenticate. A hunchsheet.app page opens in your browser: paste your key and click Connect. The key never goes into a file, and Claude Code renews the sign-in on its own.
The plugin adds the server, a skill that tells Claude when to use each tool, and three commands for CSV files:
| Command | Writes back, for each row |
|---|---|
/hunch:hunch-check <file.csv> <column> <yes/no question> | The probability of yes |
/hunch:hunch-tag <file.csv> <column> <label1,label2,...> [question] | One of your labels, with its confidence |
/hunch:hunch-score <file.csv> <column> <question> <level1,...,levelN> | A level from your scale, with its confidence |
For scripts or CI where no browser is available, add the server with the key in a header instead of installing the plugin:
claude mcp add --transport http hunch https://hunchsheet.app/mcp --header "Authorization: Bearer hunch_your_key_here"Gemini CLI
gemini extensions install https://github.com/Eddy1919/hunch-mcpGemini CLI asks for your key during install and stores it as a sensitive setting.
Claude, ChatGPT, Cursor and VS Code
- Claude (claude.ai and Claude Desktop): add a custom connector with the URL
https://hunchsheet.app/mcp, then sign in with your key when Claude opens Hunch's page. - ChatGPT: in developer mode, create a connector with the same URL and sign in the same way.
- Cursor and VS Code: one-click install buttons on hunchsheet.app/connectors.
Data
The texts and question in a tool call go to the Hunch server and to TypeSafe to compute the answer, and are not stored. Details at hunchsheet.app/privacy.
License
The files in this repo are MIT licensed. The Hunch service has its own terms.
Questions fréquentes
Qu'est-ce que Hunch ?
Calibrated judgments for text: yes/no probabilities, picks from your options, or scores.
Comment installer Hunch ?
Il fonctionne comme serveur distant. Dans Claude Code : claude mcp add --transport http hunch 'https://hunchsheet.app/mcp' --header 'Authorization: <Bearer token: "Bearer hunch_<40 hex characters>". Get a free key (100 rows) at https://hunchsheet.app.>'. Pour les autres clients, collez la configuration JSON de cette page dans leurs réglages MCP.
Faut-il une clé API pour Hunch ?
Oui. Il attend Authorization, à renseigner dans sa configuration.
Hunch est-il open source ?
Oui. Son code est sur GitHub sous licence MIT.
Qui maintient Hunch ?
Il est publié par Eddy1919. AI.tn le référence sans le maintenir ni le recommander.
Issu du registre MCP officiel. Être listé n'est pas une recommandation : vérifiez le code et les permissions d'un serveur avant de lui donner accès à vos données.