Hey Product Hunt! 👋 AI coding assistants are incredible, but context management is completely broken: 1. Multi-file questions dump massive raw files into the prompt, exhausting Claude Pro / ChatGPT Plus 5-hour rate limits by midday. 2. Teams on metered APIs bleed thousands of dollars monthly on boilerplate tokens. 3. Naive text summarizers hallucinate and strip vital types, causing code patches to break. We built **Tokonomics** to treat prompt engineering like a compiler: 🌲 **14-Language AST Pruning:** Tree-sitter WASM parses code in-memory, preserving exact types and interfaces while pruning dead bodies. ⚡ **3x–4x More Prompt Headroom:** Cuts up to 75% of context payload weight so you stop getting locked out during work hours. 💬 **Dockable Native Chat Panel:** Dedicated secondary sidebar chat + editor split view with streaming Markdown, local session history, and CLI subscription routing (`@tokonomics /claude` and `@tokonomics /codex`). 🔒 **Zero-Egress Privacy:** Runs 100% locally. Zero telemetry. No proxy servers. Available now on the Visual Studio Marketplace! Looking forward to your thoughts and bug reports.

Forum Threads

p/tokonomics-2

•

2mo ago

14-Language AST Context Compiler & AI FinOps for VS Code

Stop hitting Claude and GPT rate limits. Tokonomics parses your code with local AST parsers to prune 75% of context bloat, giving developers 3x 4x more prompt headroom with zero telemetry and a native dockable chat panel.

View all