Use Codex
Local workspace, shell, tests, git operations, release assets, and evidence collection are the hard requirements.
Free AI coding-agent workflow picker
Choose the agent that can prove the work.
Score Codex, Claude Code, Cursor, and local agents against repo control, reasoning depth, IDE context, UI QA, release proof, and handoff risk.
Routing Inputs
Pick the task shape and proof needs. The picker routes to the cheapest reliable agent that can produce direct evidence.
The primary shape of the next agent run.
Higher means file edits, tests, git, or release assets need local proof.
Higher means plan critique, broad tradeoffs, or second-opinion review matter.
Higher means inline editor context and fast local code navigation matter.
Higher means screenshots, UI state, device checks, or visual proof are required.
Higher means work crosses models, sessions, reviewers, or ownership boundaries.
Routing Note
Copy this into a run brief, handoff packet, or order-intent issue before buying the full workspace.
Decision Rules
The paid pack adds durable run logs, verification ledgers, model routing, and handoff packets around this choice.
Use Claude Code
Long-context reasoning, plan critique, architecture judgment, or adversarial review is the bottleneck.
Use Cursor
The work is fastest inside the IDE with inline code context, focused edits, and direct developer review.
Revenue Rule
Agent routing is not revenue. This page does not create checkout, prove payment, publish the paid ZIP, or count as revenue. Gross revenue is counted only from checkout, receipt, payout, or seller-dashboard evidence.