canonical: https://jentic.com/apis/hunchsheet.app/hunchsheet # Hunchsheet App Hunch API The Hunch API judges batches of short texts against a question and returns one typed answer per text instead of generated prose. Each call takes up to 40 texts and one of four kinds: a yes/no probability, a pick from up to 255 options, a position on a 2 to 10 level scale, or one probability for each of up to 10 yes/no questions. Billing is one credit per answered text, with blank texts and repeats inside a call free, and a free quote call prices a job before it runs. An agent can mint its own 100-row trial key, check the credits left on a key, and hand a person a Stripe checkout link to add more. ## For AI agents Classify, score, or yes/no-judge up to 40 short texts per call and get one typed result per text, with free quoting, balance checks, a self-serve 100-row trial key, and checkout links for more credits. ## Scope Does not generate free-form text, summarise documents, or train custom models. Use for typed judgements on batches of short texts only. ## Capabilities - Ask one yes/no question of up to 40 texts and get a probability for each - Sort texts into one of up to 255 labelled options, with confidence and per-option probabilities - Place each text on a 2 to 10 level scale such as angry to delighted - Answer up to 10 yes/no questions per text in a single multi call - Price a job before running it with the free quote call, given rows and questions - Mint a 100-row trial key for an agent and create a Stripe checkout link for a person to add credits ## Use cases ### Bulk text judgement from an agent through Jentic A developer's agent hands Hunch the repetitive part of a review: hundreds of short texts that each need the same judgement. Through Jentic the agent finds the ask operation by intent and sends rows in batches of 40; the Hunch key is injected by the developer's own Jentic One instance, and a rule can stop the agent from creating checkout links. Example prompt: Call GET `/v1/quote` with rows 120 to confirm the cost, then call POST `/v1/ask` three times with kind noul, the question 'Is this lead a decision maker who can approve a purchase without asking someone else?', and 40 rows each ### Support ticket triage Route incoming tickets by sorting them into categories. The choice kind takes a list of options, each written as a label or 'label: description', and returns the chosen label with a confidence value and a probability per option, so low-confidence tickets can go to a person. Example prompt: Call POST `/v1/ask` with kind choice, options ['billing: invoices and charges', 'refund', 'bug', 'other'], and the 40 newest ticket texts as rows, and flag any result whose conf is below 0.6 ### Review sentiment scoring Put customer reviews on a scale rather than a binary. The score kind takes 2 to 10 options ordered low to high, such as angry, disappointed, neutral, happy, and delighted, and returns a weighted position, the index and label of the most likely level, and a confidence. Example prompt: Call POST `/v1/ask` with kind score, the question 'How does the reviewer feel about the product overall?', options angry, disappointed, neutral, happy, delighted, and 40 review texts as rows ### Multi-question lead qualification Answer several yes/no questions about each text in one pass. The multi kind takes 1 to 10 questions and returns one probability per question for every row, billed one credit per question per text, so a lead can be checked for buying power, urgency, and frustration together. Example prompt: Call POST `/v1/ask` with kind multi, questions ['Can they buy?', 'Is it urgent?', 'Are they angry?'], and 40 inbound email snippets as rows, and list the rows where the first two probabilities are above 0.7 ## Key endpoints | Method | Path | Description | | --- | --- | --- | | POST | `/v1/ask` | Judge up to 40 texts | | GET | `/v1/quote` | Price a job before running it | | GET | `/v1/balance` | Credits left on this key | | POST | `/trial` | Get a free 100-row key | | POST | `/v1/checkout` | Create a buy link for a person to open | ## Key resources - **Ask jobs** — Batches of up to 40 texts judged as noul, choice, score, or multi - **Quotes** — Free price checks for a number of rows and questions - **Keys and credits** — Trial keys, credit balances, and Stripe checkout links for Starter or Top-up plans ## AI readiness This API is usable in Jentic One now. Its AI-readiness score against Jentic's framework shows where it stands today and where improvements would make it even easier for agents to use. - **Score:** 67 / 100 - **Maturity:** AI-Aware - **Dimensions:** - Foundational Compliance: 100 / 100 - Developer Experience & Jentic Compatibility: 64 / 100 - AI-Readiness & Agent Experience: 50 / 100 - Agent Usability: 94 / 100 - Security: 60 / 100 - AI Discoverability: 68 / 100 - **View full report:** https://jentic.com/apis/hunchsheet.app/hunchsheet/scorecard - **How the score is calculated:** https://docs.jentic.com/reference/api-readiness-framework/overview/ - **More about the dimensions:** https://docs.jentic.com/reference/api-readiness-framework/specification/#dimensional-model-overview ### Score it yourself Every API in the directory is allowlisted, so you can re-score it with no key required. - **Score your own API:** https://jentic.com/scorecard.md - **Scoring CLI agent skill:** https://github.com/jentic/jentic-api-scorecard/blob/main/skills/jentic-api-scorecard/SKILL.md ```sh npx @jentic/api-scorecard-cli score ``` ## Why Jentic - **Setup:** Wiring the Hunch API by hand means minting or buying a key, sending it as a bearer token, batching rows into groups of 40, and handling the out-of-credits case. Through Jentic you install once, import Hunch from the API Directory, store the key once, and your agent calls it. - **Permission scoping:** Hunch carries the texts and the question inside the request body, so rules bound which operations your agent may call, not what it asks. You choose the operations it may call, so trial key minting and checkout link creation are not included unless you add them. - **Credential handling:** Your Hunch key is stored once, encrypted, by your own Jentic One instance and injected at execution time. It never enters the agent's prompt, logs, or context. - **Discovery method:** Agents search Jentic by intent such as 'classify short texts' or 'score reviews on a scale', and Jentic returns the matching Hunch operation with its input schema so the agent calls the right endpoint without browsing the reference docs. ## Related APIs - **OpenAI API** — General language models you can prompt to classify text yourself - **Cohere API** — Language models with embeddings and reranking for text tasks - **Zendesk API** — Support tickets to classify and route ## FAQ ### What authentication does the Hunch API use? Per its OpenAPI spec, the Hunch API uses a Hunch key sent as a bearer token, in the form hunch_ followed by 40 hex characters. POST /trial needs no key and returns one with 100 free rows, and GET `/v1/quote` works with or without a key. Through Jentic the key is stored encrypted by your own Jentic One instance and injected per call. ### How is the Hunch API priced? Hunch's spec states one credit per answered text, with blank texts and repeats inside a call free, and one credit per question per text for the multi kind. After the 100 free trial rows, the spec lists $29 for 5,000 rows and $19 for 25,000 more, with no subscription. GET `/v1/quote` prices a job for free before you run it. ### What are the rate limits for the Hunch API? The OpenAPI spec does not specify request rate limits for the ask endpoint; each call is capped at 40 rows, and POST /trial is limited to 2 keys per address per day. Check https://hunchsheet.app for current limits. ### What happens when a Hunch key runs out of credits? By default POST `/v1/ask` returns 200 with per-row errors. If you send the optional X-Hunch-Agent header with your agent or product name, it instead answers 402 with a checkout offer. POST `/v1/checkout` returns a Stripe link that a person must open and pay; the spec states an agent cannot pay for them. ### How do I classify texts with the Hunch API through Jentic? Search Jentic for 'classify short texts' to get POST `/v1/ask`, then send kind choice with your options and up to 40 rows, splitting larger sets across calls. To run it on your own infrastructure, install Jentic One from its GitHub repo and store your Hunch key once. ### Is there a Hunch MCP server? Yes. Hunch's own spec says the same engine is available as an MCP server. You can also connect the Hunch API directly through Jentic: your agent gets the full 5-endpoint surface without loading another server's tool definitions into its context, your key is injected at call time by your self-hosted instance instead of living in an MCP config, and your rules decide which operations the agent may call. ### Can I limit what my agent is allowed to do with the Hunch API? Yes. For example, write a rule that allows only GET `/v1/quote`, GET `/v1/balance`, and POST `/v1/ask`, so the agent can price, run, and track jobs but cannot mint trial keys or create checkout links. Every call it makes is logged by your Jentic One instance.