Hi Product Hunt, My name is Dan Levitan. I built ICC-GO because LLM work often starts as a chat, but serious AI work needs a record: what model ran, what constraints applied, what previous output was used, and what file was produced. ICC-GO is a local-first notebook for Intent-Cell Coding. A cell can declare routing, limits, references, and file output directly in the prompt: > claude.max < latency <= 1m @forward c3 %from c1 The goal is simple: make AI work survive the chat window. What is different: - model routing is visible per cell - constraints live beside the prompt - cells can reference earlier cells with %from - outputs can become files, not just chat text - workflows can be inspected, rerun, and shared - the local app is BYOK ICC-GO is early, but the local bundle, examples, language docs, and public repo are ready to inspect. I would love feedback from developers who build repeatable AI workflows, agent pipelines, research notebooks, or internal LLM tools.

ICC-GO: AI work should survive

Local-first notebooks for executable LLM workflows.

Local-first notebooks for executable LLM workflows.