I built an open-source tool that cuts codebase context by 95% using AST pruning — offline, free, works natively in Cursor

Showcase Built for Cursor
, ,
1

I built T-Zero Context Architect out of frustration with AI coding agents (Cursor, Claude Code, Cline) constantly exhausting their context windows and token budgets on medium-to-large repositories.

Most tools either dump raw source files into the prompt (blowing up token limits) or use lossy embeddings that hallucinate missing function signatures.

T-Zero takes a different approach:

1. AST Skeletonization: It parses Python, JS/TS, Go, Rust, Java, and C++ into lightweight structural outlines (signatures, docstrings, classes, interfaces) while stripping function bodies. A 166k token codebase becomes 8.4k tokens (-94.9% reduction).

2. Zero-Leak Security: Uses OS-level keyrings (Windows Credential Manager, macOS Keychain, Linux Secret Service). Zero telemetry, no cloud embedding APIs required.

3. Native MCP Server: Exposes 21 tools over stdio for autonomous AI agents in Cursor, Claude Desktop, Antigravity IDE, and Windsurf.

4. $0 Cost: 100% offline, reproducible locally via `uvx tzero-mcp` or `python tzero.py –savings`.

GitHub: GitHub - toprakahmetaydogmus/TZeroAlgorithm: AST Codebase Context Engine & 21-Tool MCP Server for AI Coding Agents — Up to 95% Token Reduction & Zero-Leak Security Architecture · GitHub

I would love to hear your feedback on the AST pruning heuristics and what languages/features you would like to see next!

4

Any plans to support Swift and Kotlin, for us mobile app developers?

5

ım developing what you say right now. thx for idea

6

Hey @RayGraham ! :waving_hand:

Great news — your request is officially shipped! :rocket:

We just released T-Zero v4.1.0 (Mobile Edition) with full native AST signature extraction and context pruning specifically built for mobile developers:

:mobile_phone: What’s inside for Mobile Devs:

  • Swift (.swift) AST Engine: Preserves protocols, structs, classes, actors, extensions, and function/init contracts. It also keeps your SwiftUI property wrappers (@State, @Binding, @Observable, @MainActor, @Published, @EnvironmentObject) intact while stripping execution bodies — slashing token overhead by 60% to 80% on iOS codebases.
  • Kotlin (.kt, .kts) AST Engine: Deep interface outliner for data classes, sealed interfaces, companion objects, and @Composable functions, removing internal loop/logic noise while maintaining complete API signatures.
  • Dedicated Mobile Presets: Quick-scoped workspace modes for Mobile (iOS Swift) and Mobile (Android Kotlin).
  • Polyglot Mobile Dependency Graph: Mobile imports are now tracked across module dependency trees and MCP tools.
  • Zero Cost & 100% Offline: Runs completely locally with $0.00 cloud costs and 100% Zero-Leak security.

Prebuilt standalone Windows binaries (TZeroAlgorithm.exe, TZeroMCP.exe, AddMCP.exe) and checksums are available on the release page:
:backhand_index_pointing_right: T-Zero v4.1.0 Release on GitHub

Give it a spin on your mobile project, and if you find it helpful, dropping a :star: on the GitHub Repository would mean the world to us! :glowing_star:

7

Wow, amazing! So cool, thanks!

9

can ı get a star my brother ?

11

Done!