7 Platforms People Use to Build Jev Workflows

A model that returns decisions still needs somewhere to run. Jev answers closed questions with a choice, a score, or a yes/no probability, but it doesn't schedule anything, listen for new tickets, or act on its own answers. Application code decides when to call it and what to do with the result. That means every Jev workflow lives on some platform, and the options are scattered across TypeSafe's own tools, cloud providers, automation apps, and community projects.

Picking one means comparing tools that do very different jobs, so this guide takes them one at a time, says how each plugs in, and says who each suits. This guide covers what a Jev workflow needs from a platform, seven platforms people use to build Jev workflows, and how to choose between them.

• A platform around it: Jev returns a decision for each question and leaves triggers, data, and actions to the platform it runs in.
• Seven platforms: PromptQL, Zapier, n8n, Make, TypeSafe's API and SDKs, Vercel AI Gateway and AI SDK, and LangChain with LangGraph and LangSmith.
• No usage figures: The list covers platforms with official or provider-documented integrations and published builds, not a popularity ranking.
• Pick by how your team already works: The platform already in use usually makes the best first home for a Jev step.

Jev handles a single step in a workflow. It takes some text and a set of questions and returns typed answers with probabilities. Everything around that step, such as what starts the workflow, what happens to each answer, and who sees the result, has to come from somewhere else. That somewhere else is the platform, and the same five jobs show up whichever platform a team picks:

• A trigger or data source: Something starts the workflow, such as a new ticket, an incoming email, or a table of records.
• A place to call Jev: An app, function, or integration sends the text and the questions and receives the answers.
• Branching on the answer: Rules, usually thresholds on the probabilities, decide what happens next.
• A handoff for uncertain items: A route to a person, or to a stronger model, takes over when confidence is low.
• Logging: A record of the input, the answers, and what was done lets decisions be reviewed later.

Each platform below covers some of these jobs and leaves the rest to the builder.

No platform publishes usage figures, so this list covers seven where Jev has an official or provider-documented integration and builders have published templates, tutorials, or workflows.

1. PromptQL

PromptQL is an AI workspace where teams work with their data in shared threads. It connects to warehouses, databases, and SaaS apps.

• How Jev plugs in: In workspaces where Jev has been provisioned, a plain-English request hands the repeated decisions in a task to Jev, such as sorting the last seven days of support tickets, as using Jev without writing code lays out. Results come back in the thread with a confidence score on each answer, so low-confidence items can be reviewed first.
• Best for: Batch decisions over company data, such as sorting a backlog, labeling a large set, or joining labels to account or revenue data. Teammates can correct the criteria in the thread, then the setup is reused or turned into a dashboard.
• Worth knowing: It isn't a trigger-based automation builder or the live request path. Decisions inside an application are called from the application or from one of the platforms below.

2. Zapier

Zapier is a no-code automation platform that connects apps through triggers and actions.

• How Jev plugs in: Zapier has an official TypeSafe Jev app with an Ask Questions action. It handles yes/no, pick-one, and scale questions about text from any trigger, such as a Gmail email, a form response, or an Airtable row, and lets you set a minimum confidence.
• Best for: Operations teams sorting email, tickets, and leads without writing code.
• Worth knowing: Zapier handles your data in transit, so check its policies as well as TypeSafe's.

3. n8n

n8n is a workflow automation platform with a visual canvas, available as a cloud service or self-hosted.

• How Jev plugs in: An official TypeSafe AI node, built by TypeSafe and verified by n8n, documents Evaluate and Route capabilities that return a choice and a confidence value a workflow can branch on. On n8n Cloud, Jev runs on Gateway credits, with no TypeSafe account or key needed.
• Best for: Teams that want confidence-based routing in a workflow tool they can self-host.
• Worth knowing: Community nodes also exist, and some, such as one community package, are not verified for n8n Cloud, so the official node is the safer default.

4. Make

Make is a visual automation platform where workflows are built as scenarios of connected modules.

• How Jev plugs in: Make's official TypeSafe app has an Evaluate a state module, which judges content against the typed questions you write and returns structured answers with probabilities and confidence, and a raw Make an API call module. It connects with a TypeSafe API key.
• Best for: Teams already building scenarios in Make who want a decision step that branches on the answer.
• Worth knowing: It requires a TypeSafe account, and Make's template gallery includes TypeSafe templates to start from.

5. TypeSafe's API and SDKs

TypeSafe's own platform is the direct route to Jev, with no intermediary.

• How Jev plugs in: Code sends a model, a state, and a map of typed questions to the System One API at api.typesafe.ai/v1/systemone, as a community guide shows. Official SDKs wrap it for TypeScript (@typesafe-ai/sdk) and Python (typesafe-sdk), and the Playground lets anyone test questions in a browser without code.
• Best for: Developers wiring Jev into their own services.
• Worth knowing: The trigger, thresholds, retries, and logging all stay in your code.

6. Vercel AI Gateway and AI SDK

Vercel's AI Gateway serves models from many providers through one account, and the AI SDK is its TypeScript toolkit.

• How Jev plugs in: Jev is available on the gateway as typesafe-ai/jev, and the AI SDK exposes it through its experimental evaluate API, which takes a state and a map of questions. Vercel's own example turns a support case into a queue, a priority, and a refund-review decision, and sends uncertain routing to manual review.
• Best for: TypeScript teams already building on Vercel.
• Worth knowing: The interface differs from TypeSafe's own, and the Boolean question maps to Noul. Vercel's example also turns on zero data retention and no training per request.

7. LangChain, LangGraph, and LangSmith

LangChain is a framework for building agents and LLM pipelines, with LangGraph for stateful workflows and LangSmith for evaluation and tracing.

• How Jev plugs in: LangSmith Evals can use Jev as a judge, and community examples show Jev inside LangChain and LangGraph workflows for support triage, model routing, and guardrails.
• Best for: Teams building agents that need a routing, guardrail, or judging step.
• Worth knowing: LangChain's Jev middleware is still experimental. The LangSmith test is LangChain's own, and LangChain is an integration partner, so results are worth confirming on your own cases.

A Jev workflow is a model plus a platform, and the right platform depends on how a team already works: a no-code tool for automations, an SDK or gateway for code, a framework for agents, or an AI workspace for batch work over company data. With no usage counts to lean on, the useful test is fit. The platform should be able to branch on a probability, hand uncertain items to a person, and log what happened. Starting with a platform already in use keeps the new part small, which is one decision step.

Which platform should you start with?

Start with the one that matches how you already work. If you don't code, use Zapier, n8n, or Make. If you write code, use the TypeSafe SDK, or Vercel's AI SDK if you build in TypeScript on Vercel. If you build agents or evaluate LLM outputs, use LangChain, LangGraph, or LangSmith. If the decisions run over stored company data, use PromptQL.

Is there a visual way to build Jev workflows?

Yes. Zapier, n8n, and Make are visual builders with official Jev integrations. A community demo built with Liveblocks also shows a multiplayer workflow builder for Jev, but it is a reference demo, not a product.

Which integrations are official?

TypeSafe's own API, SDKs, and Playground are official, and Zapier, Make, and n8n each have official Jev integrations. Vercel documents Jev on its AI Gateway and AI SDK, and LangChain documents Jev in LangSmith Evals. Community nodes for n8n, framework examples, and demos are built by independent developers.

Does the host change how you call Jev?

Yes. Model IDs and request shapes differ. TypeSafe's own API uses the System One endpoint, while Vercel exposes typesafe-ai/jev through the AI SDK's evaluate API, where the Boolean question maps to Noul. Other gateways such as Cloudflare Workers AI and OpenRouter also serve Jev, each with its own model ID. The host also handles data in transit, so check its terms.

Where can you find examples other people have built?

madewithjev.com is a community directory of builds, listing the cost, latency, and source each author reported. It isn't an official TypeSafe site, so treat the figures as author-reported.