Topos – Structural code quality metrics for agent-written programs

Name: Topos – Structural code quality metrics for agent-written programs
Availability: InStock
Author: wayland_jeremy

by wayland_jeremy·Jun 25, 2026·1 point·1 comment

Visit Project View on HN

AI Analysis

●●SolidBig BrainSolve My Problem

Heyting algebra-based quality medals beat binary pass/fail tests for AI code.

Strengths

•Graph-based structural analysis (AST, CFG, CPG, MDG) catches architecture issues tests miss
•MCP server integration works with Claude Code, Cursor, Gemini CLI out of the box
•Three orthogonal pillars (Simple, Composable, Secure) give agents concrete structural targets

Weaknesses

•Code quality metrics is a crowded space with SonarQube, CodeClimate already established
•Category theory framing may alienate practitioners who just want actionable feedback

Post Description

Agents generate code faster than humans can review it. The problem isn't correctness — passing tests only prove code ran, not that it belongs in the codebase. Topos parses programs into graphs (AST, CFG, CPG, MDG) and scores three orthogonal structural pillars — Simple, Composable, Secure — awarding a Quality Medal per file. The evaluation is genuinely a topos (category theory): our subobject classifier is the Heyting algebra freely generated by {Simple, Composable, Secure}, whose elements are exactly the medal states agents navigate. Open source. CLI + MCP server (Claude Code, Cursor, Gemini CLI, Windsurf) + VS Code extension. GitHub: https://github.com/Krv-Labs/topos