How Can You Use Jev Without Writing Code?

Five hundred support tickets sit in a spreadsheet, and one question needs an answer. Which ones are urgent? Most guides to Jev open with API keys, SDKs and JSON, which makes the whole thing sound like a job for an engineer.

Good news for anyone who does not code. Jev AI is TypeSafe AI's decision model, and it answers yes/no, pick-one and scale questions about text, with a confidence score on each answer. You can use Jev without writing code through TypeSafe's Playground, official apps for Zapier, Make and n8n, or a plain-English request in PromptQL.

• The TypeSafe Playground is the quickest way to test your questions before you automate anything.
• Zapier, Make and n8n all have official Jev integrations, so you can run Jev on live data with no code.
• PromptQL runs Jev from a plain-English request, with the results shared in a thread.
• Results depend on how clearly you write each question and its answer options.

There are four ways to use Jev without writing code, and PromptQL adds a fifth. Zapier, Make and n8n have official integrations, and the TypeSafe Playground is TypeSafe's own web tool.

1. TypeSafe Playground

Source

The Playground is TypeSafe's browser tool for testing Jev before you automate anything. You log in, paste your text in as the state, add your questions and see every answer at once. TypeSafe's quickstart guide walks through it with a customer support ticket.

Here is what each question type returns:

• Noul (yes/no): A single probability from 0.0 to 1.0. Asking "Does this message express urgency?" can return 1.0.
• Choice (pick one): The selected option, a confidence score and a probability for every option.
• Score (rating scale): A numeric score, a confidence level and a probability for each point on the scale.

The quickstart example runs three questions on one ticket. It routes the ticket to billing, sales or technical, rates how frustrated the customer is and checks for urgency. You can build the same kind of multi-question test with your own records.

The Playground suits one-off checks. Once your questions give answers you trust, move to one of the next routes to run records automatically.

Pick this if: you want to check that your questions and options work before you build anything.

2. Zapier

Source

Zapier has an official TypeSafe Jev app with an "Ask Questions" action. You add it to a Zap as a step and send it text from any trigger, such as a Gmail email, a form response or an Airtable row.

What you set up in the action:

• Content: The text Jev should judge. This is the only required field.
• Questions: Yes/no questions, pick-one questions and scale questions, each with its own name and wording.
• Safeguards: A minimum confidence threshold and a setting for how to handle answers Jev is unsure about.

Each answer comes back with a confidence score. You can then use Zapier's filters and paths to send each item where it belongs, such as urgent messages to a team channel and low-priority ones to a spreadsheet.

Zapier connects to over 9,000 apps, so the text can come from almost anywhere your team already works.

Where it fits best: teams that already run their workflows in Zapier and want Jev to screen incoming messages for topic, urgency or tone.

3. Make

Source

Make has an official TypeSafe app with an "Evaluate a state" module. The module judges your content against the typed questions you write and returns structured answers with probabilities and confidence.

What to know about the app:

• Modules: "Evaluate a state" does the judging. The "Make an API call" module is a raw option you can skip.
• Connection: Log in to TypeSafe, copy your key from the API keys page, then add a TypeSafe module in Make, choose "Create a connection" and paste the key.
• Templates: Make's template gallery has ready-made scenarios, so you can start from a working example.

Make's routers split a scenario into branches, which suits flows where the next step depends on Jev's answer. For example, one branch can handle high-confidence "Urgent" results automatically while another collects uncertain ones for review.

Make's documentation for the app was last updated on September 25, 2026, so the setup details are current.

Who it suits: people who like to build visual, branching scenarios and are comfortable creating an API key once.

4. n8n

Source

n8n has an official TypeSafe AI node, built by TypeSafe and verified by n8n. n8n describes it as a smart If node for fuzzy decisions, since it returns probabilities for each outcome and never writes text.

What to know about the node:

• Operations: Evaluate returns Jev's answers, and Route sends each item down its own named output, with one route per outcome.
• Confidence control: Every answer carries a confidence score, so you can auto-route only above a threshold, such as 95 percent, and send the rest to review.
• Setup: On n8n Cloud, Jev runs on Gateway credits with no TypeSafe account or key. It is free until October 10, 2026 at 23:59 UTC, then standard Gateway credit rates apply. Self-hosted setups bring their own key, and an instance owner installs the verified node first.

Typical uses include sorting support tickets by priority and screening leads or form submissions, and a Gmail and Google Sheets triage template gives you a working example to copy. Regular If nodes still suit fixed rules, and chat models suit anything that needs written text. Community-built Jev nodes also exist, but the official node is the safer start because TypeSafe maintains it.

Ideal if: you want to try Jev on n8n Cloud without a TypeSafe account or key.

Source

You use Jev in PromptQL by describing the decision in plain English and letting the workspace run it across your records. Jev works in workspaces where it has been provisioned, so the first step is to check.

Follow these steps:

1. Open a PromptQL workspace: Sign in, create a project and start a bot.
2. Connect your data: Link the source where your records live, or describe the records in the thread.
3. Check that Jev is available: Ask, "Is Jev available in this workspace? If not, tell me before we start."
4. Give your criteria in plain English: Share the background to judge against, the labels you want back, your deal-breakers and your edge cases.
5. Ask PromptQL to run Jev across the set: Say how many records to include, which labels to use, what to flag and what to do if Jev is unavailable.
6. Review the output in the thread: You get a ranked shortlist, the reasons and a list of low-confidence items for a human look.
7. Share and reuse it: Invite teammates into the thread to correct the criteria, then reuse the setup for the next batch or turn the result into a dashboard.

Here is a ready-to-copy prompt for sorting support tickets:

Is Jev available in this workspace? If not, tell me before we start.If it is, use Jev to sort the support tickets from the last 7 days in the Support table. We run a project management app for small agencies. Label each ticket Urgent, Normal or Low based on whether the customer is blocked from working. Treat billing errors as Urgent. Add an Other label for anything unclear. List the Urgent tickets first with a reason for each, then list every ticket where Jev's confidence is low so a person can check it. If Jev is not available, stop and tell me.

When a teammate corrects a rule in the thread, ask PromptQL to rerun the batch and compare the two results. To see how Jev pairs with a chat model for the parts that need writing, read how to build an LLM workflow with Jev. For another worked example of the same plain-English approach, see how to use AI for your job search.

Four things matter before you start.

• Cost: Jev is priced at $0.042 per million input tokens, and output is free.
• Access: TypeSafe reopened signups on September 27, 2026, and new accounts get no free credits. This was checked on October 3, 2026.
• Trust: Test on a labeled sample first, use the confidence score to decide what runs automatically and send low-confidence items to a person.
• Fit: Jev handles closed questions over text. Anything that needs a written explanation belongs with an LLM, and this comparison of Jev and general LLMs shows where to draw the line.

Whichever route you pick, the question you give Jev decides the quality of the results. Before you open anything, fill in a one-page question card.

A good card holds:

• The question: One clear thing to decide, such as whether a customer is blocked.
• The options: Each answer with a one-line definition, plus an Other option for unclear cases.
• Known examples: Three records per option where you already know the right answer.

Paste the card into the Playground, a Zap, a Make scenario, an n8n node or a PromptQL thread. If Jev matches your known examples, you are ready to run the full set.

Can Jev Read Images, Audio or PDFs?

No. Jev accepts text or JSON only, and its primary training language is English. Convert images, audio and PDFs to text first, then send the text.

Can Jev Explain Why It Chose an Answer?

No. Jev returns an answer and a confidence score, and it never writes text. If you need a written reason, pass the item to a chat model afterward, as covered in this guide to Jev versus ChatGPT.

Can You Use Jev Inside Google Sheets or Excel?

TypeSafe's docs list no official Sheets or Excel add-on. Community-built add-ons exist, so review any add-on before you send company data through it. Zapier and Make can also read rows from a spreadsheet and send them to Jev.

Is Your Data Sent to TypeSafe When You Use Jev?

Yes, the text you send goes to TypeSafe, which hosts Jev in the US. According to TypeSafe's privacy policy, it does not train or fine-tune on your input and does not disclose it to third parties other than service providers. Zapier, Make and n8n also handle your data in transit, so check their policies too.

1. PromptQL Blog: The latest best from PromptQL! - promptql.io
2. PostgreSQL: Documentation: 18: 5.9. Row Security Policies - www.postgresql.org