arXivarchitecture2026-01-30

Agentic Much? Adoption of Coding Agents on GitHub

Romain Robbes, Théo Matricon, Thomas Degueule, Andre Hora, Stefano Zacchiroli

View Original →

Abstract

In the first half of 2025, coding agents emerged as a category of development tools that rapidly transitioned from research to practice. Unlike traditional code completion LLMs such as Copilot, agents like Cursor, Claude Code, and Codex operate with high degrees of autonomy — generating complete pull requests from developer-provided task descriptions. This paper presents the first large-scale study (129,134 projects) of coding agent adoption on GitHub, finding an estimated adoption rate of 15.85–22.60%, which is very high for a technology only a few months old — and still increasing.

Relevance to Tokalator

This empirical evidence directly motivates Tokalator's existence: as agentic coding becomes mainstream, token budget visibility becomes critical. Agentic sessions consume far more tokens per interaction than traditional code completions, making real-time budget monitoring (Tokalator's core feature) increasingly valuable.

cs.SE

Related Articles