Which AI coding tool do practitioners recommend right now?

Claude Code remains the most recommended AI coding tool, backed by numerous corroborated practices and practitioner anecdotes. OpenAI Codex is a strong second, with significant internal use at OpenAI and emerging integrations. New preferences have emerged for Cursor Composer 2.5 for token efficiency and Lovable for app generation, though adoption appears narrower. Hermes Agent has high corroboration for autonomous workflow execution.[9]

While Claude Code continues to dominate, Cursor ⁵²³ and Lovable ²²¹ have been preferred for specific use cases in recent claims. Hermes ³²¹ has emerged as a highly corroborated multi-agent tool, though it is not a direct coding tool. This adds diversity but does not change the leading position.[3]

Claude Code: Widely used and recommended for coding, with practices like using loops ⁵¹⁸ and structured workflows ¹²¹ to achieve rapid results, such as building websites in 18 minutes ¹³² and profitable trading bots ⁴⁴⁶. Power users leverage CLAUDE.md for context ¹⁷⁸ and multi-agent setups ³⁷⁶.[6]

OpenAI Codex: Used by 90% of OpenAI engineers ¹¹⁵ and capable of handling 40% of coding tasks via agentic loops. It can be integrated with Claude Code for multi-agent collaboration ¹⁴¹ and offers /dynamic skill for parallel task execution ⁵⁷¹.[3]

Cursor: Cursor Composer 2.5 is preferred over alternatives for token efficiency on the Pro+ plan ⁵²³.[1]

Lovable: Lovable is preferred over Claude Code and Cursor for generating complete apps with design, logic, backend, and deployment from natural language ²²¹.[1]

Hermes Agent: Highly corroborated for learning and autonomously executing repetitive workflows after a single demonstration ³²¹. Its judgment system enforces completion contracts ⁵⁷².[2]

Other tools: Orca enables parallel agent teams ¹³⁷, Dotcode provides browser-native coding sandboxes at low cost ¹³⁰, and Fable can one-shot CLI agents ¹³⁵.[3]