Prompteus

Maker

📌

👋 Hey Product Hunt! I’m Bap, Co-founder at Prompteus.

I’m thrilled to introduce Prompteus — a complete solution to build, deploy, and scale AI workflows.

Over the past few years, I’ve built products with my team for governments and Fortune 500s in highly regulated industries. When LLMs exploded, we started integrating them into everything — and the same pain points kept popping up:

How do I log requests and track costs?
How do I switch models without rewriting my app?
How can I make sure the response never says a specific word?

That’s why we built Prompteus.

Instead of hardcoding AI calls all over your stack, Prompteus gives you a visual workflow editor to design, manage, and deploy AI logic — no infra, no spaghetti prompts, no DevOps overhead.

We call these workflows Neurons. You drag and drop building blocks (like model calls, conditionals, and transformations) and deploy them as secure API endpoints. They work from your frontend, backend, anywhere. You can even chain Neurons together.

✨ Highlights:

Multi-provider orchestration (OpenAI, Anthropic, Mistral, Cohere, Google…). Change models without changing a line of code.
Adaptive semantic caching to skip redundant LLM calls. Save on execution time and request cost!
Built-in auth, rate limiting, access controls — call your Neurons from your frontend if you'd like.
Detailed per-request logs, and cost analysis down to the microdollar (yup, we had to come up with that one!)
Powerful guardrails: catch input/output patterns before they hit the model. A great use case is to remove sensitive information before AI calls.

We’ve designed Prompteus so non-devs can contribute. No YAML. No redeploying a whole project for every config tweak.

There’s a generous forever free tier to get started, and we’re already testing some cool new features like tool calling, MCP server support, and more with early users.

Check out the docs, watch our videos, try it out, and tell us what you think. We’d love your feedback — AMA below!

— bap & the Prompteus team

Report

2mo ago

Min Liu

@baptistelaget Great, we are working on our own product Summizer（ https://www.producthunt.com/products/summizer ）During the development process, we do face the issue of multiple model access and switching, and we will try to address it

Report

2mo ago

Baptiste Laget

Prompteus

Maker

@min_liu7 Easily switching between models, especially if one becomes unavailable, was definitely one of the first drivers for building Prompteus!

Report

2mo ago

Jeremiah Johnson

@baptistelaget Like Zapier but for Ai ? Cool!

Report

2mo ago

Jason Yu

HabitGo

@lucasjolley_cloudraker @linjrm 🚀 This is the kind of product I wish existed a year ago.

Every time we built something with LLMs, the same DevOps nightmares came back — caching logic, model-switch rewrites, prompt spaghetti. Prompteus feels like the missing infrastructure layer between raw LLM APIs and scalable AI apps.

🔧 The idea of Neurons is smart — visual, composable, and production-ready.

💡 The semantic caching + guardrails combo alone can probably save a fortune and avoid a PR disaster.

💬 Curious to know how flexible the input/output validation is. Regex? Embedding-based? Some examples would be awesome.

This is not just another wrapper. It's Zapier meets LangChain meets Postman. Great work team — following the roadmap closely! 🔍

Report

2mo ago

Baptiste Laget

Prompteus

Maker

Thanks @kui_jason!
I think you nailed the description with Zapier meets LangChain meets Postman 😉

To answer your question about input/output validation, we do support string matching and RegEx, but the versatility of Neurons allows you to compose them together in a pretty complex way.

If you want to go a bit further, you could use one model to evaluate an input ("Does this request include medical topics", "Does this message include financial advice", etc.), and depending on that first evaluation, do different things in the rest of the workflow (or, block the execution).

As an example: our Features Deep Dive video also has an example where I go through "call summarization", removing sensitive information from the input before sending it to the LLM. Also, some useful docs on conditionals and call Neuron (from another Neuron).

Thanks for your support!