AskCodi
Custom LLMs, without training. OpenAI-compatible API.
Start new thread
Shreyans Bhansali

AskCodi - Custom LLMs, without training. Use via openai compatible api

by
AskCodi
•
We provide an OpenAI-compatible orchestration layer that lets teams compose their own “virtual models” on top of any LLM - combining prompts, reasoning, review, and guardrails, and use them everywhere from your IDE to your backend. Key features 🔌 One OpenAI-compatible API for many LLMs 🧱 Custom models you name & reuse 🧠 Reasoning mode on demand ✅ Built-in review mode 🛡️ Guardrails & PII masking 🧑‍💻 IDE & CLI integrations 📊 Analytics & cost controls

Add a comment

Replies

Best
Shreyans Bhansali
AskCodi
Maker
📌
Hey Product Hunt! 👋 I'm Shreyans, cofounder of AskCodi. After years of building askcodi as an frontend layer for constant AI releases. We are moving away from consumer tech and focusing on our expertise bring Research to Consumers. First step building an LLM orchestration layer that lets you build your own custom coding models on top of any LLM and making it accessible using openai compatible api. Instead of calling gpt-5.1 directly, you call things like: - code-generator-model - bugfix-with-review - secure-code-review Each “custom model” is a recipe: prompts + reasoning + review + guardrails. - Stack prompts, tools, reasoning, review, and guardrails into a single model name you can call anywhere. - Enable reasoning and self-check flows for complex coding tasks, even on non-reasoning base models. - Add an automatic review pass: generate, then review for bugs, security, and style in one call. - Mask sensitive data and enforce organization policies at the model layer. Our goal is simple: make small and open-source models as effective as frontier LLMs at a fraction of the cost. If you’re experimenting with AI in your dev workflow or already rely on tools like Codex, CC, Copilot, or Cursor. I’d love your feedback and support.
Masum Parvej

@shreyans_assistiv How do you manage complex reasoning on models without inherent reasoning capabilities? Curious about this part.

Shreyans Bhansali
AskCodi
Maker

@masump Thank you for your interest, we do this by making the custom model agentic which adds a reasoning feedback loop

Chilarai M
Swytchcode

@shreyans_assistiv Congrats on the launch and the great pivot! I was actually wondering how you’re offering a different product now compared to the last time I checked you guys out. Best of luck

Shreyans Bhansali
AskCodi
Maker

@chilarai Thank you for checking us out again, we have gotten out of the tool business and focusing more on research aspect of making LLMs the best personalised coding models. Our main feature is custom models which let you define model, prompts, reasoning, review and guardrails once and use it across multiple tools.

Zoya Tamboli

@shreyans_assistiv Can I add my custom memory layer or integrate my database?

Shreyans Bhansali
AskCodi
Maker

@zoya_tamboli we are working on a memory layer but not custom. Interesting suggestion, I am guessing mcp support will solve for both custom memory layer and database integration?

Daniel Baum

@shreyans_assistiv This is really sick. Given model costs, I'd love for an open source model to be as effective as frontier. How does pricing work?

Shreyans Bhansali
AskCodi
Maker

@daniel_baum Thank you for the support Daniel, it's API pricing so you pay for the tokens you consume. We also have a very generous Free tier which more than 6 free models, ever evolving.

Chimeremeze
@shreyans_assistiv ! Congrats on the Product Hunt launch 🎉 I noticed you’re building an API product. I specialize in API documentation if you ever need help with onboarding guides, API docs, examples, or reference writing, I’d be happy to support your team!
Sachin Sharma
AskCodi
Maker

Hey everyone! 👋

I’m one half of the AskCodi founding team, and I wanted to share a quick personal angle on why this launch matters to me.

Over the past few months, I’ve been using AskCodi with Cline daily for my own dev work. I’d been relying on another agentic app/CLI before, but it started slowing me down, rate limits, sluggish generations, and a drop in quality as my projects got more complex.

AskCodi changed that. Full control over prompt engineering let me finally get reliable, high-quality outputs. And with AskCodi’s orchestration layer, I could:

  • plan with a smaller, cheaper model,

  • switch to a stronger one for execution,

  • and keep it all wrapped in one custom setup.

Plus, with how fast new LLMs are coming out, AskCodi makes it dead simple to swap them in without reworking your whole workflow. That alone has saved me a ton of time.

This launch is special to me not just as a builder, but as someone who’s shipped real work using this system. If you're into agentic workflows or just want more control than typical coding assistants offer, I’d love to hear what you think. 🚀

Thanks for checking us out!

Abdul Rehman
Triforce Todos

Amazing work, Shreyans! Saving time while improving code quality is exactly what developers need.

Shreyans Bhansali
AskCodi
Maker

@abod_rehman Thank you, we have some amazing R&D to release over coming months.

Viktor Shumylo

Congrats on the launch! Love the idea of turning coding workflows into callable custom models. How you handle model evaluation across different LLMs?

Shreyans Bhansali
AskCodi
Maker
@vik_sh Thank you 🙏 we have our own set of benchmarks and have a curated of set of really tricky problems helps with a quick eval
Huisong Li
Epsilla (YC S23)

Congratulations. And happy product launch. @sachinsharma

Shreyans Bhansali
AskCodi
Maker
@huisong_li thank you 🙏
Sachin Sharma
AskCodi
Maker

Thank you !! @huisong_li

Aditya Barde

Congratulations 🎉

Great work!

Shreyans Bhansali
AskCodi
Maker
@aditya_barde Thank you for the support 🙏
Ankur Shrivastava
Moj

Interesting

Shreyans Bhansali
AskCodi
Maker

@shrivastava Thank you :)

Lilou Lane

Love this direction. Purpose-built coding “models” with built-in review and guardrails are exactly what dev teams need right now.

Mykyta Semenov 🇺🇦🇳🇱

Interesting! Is it possible to remove ChatGPT’s restrictions? It currently limits certain topics, and I don’t need such restrictions in my project.

Christophe HK
Long time user, long time happy user. what these guys do is epic, they have been able to switch from a useful Jarvis product to a robust Codi product, always integrating latest LLMs and always considering their first users. we just have to ask...Codi! Congrats for this new launch, guys. Sky is your limit!