Trae

Trae is an AI coding assistant built around agentic SOLO workflows: instead of only completing a line of code, it aims to take a bounded task, produce an implementation, and let the developer review the result. The fetched homepage was mostly a JavaScript shell, but the public pricing page returned useful static loader data and FAQ headings. That pricing evidence is stronger than the previous record, so this profile now treats core plan data as verified from curl while still advising buyers to confirm billing-period labels in the live UI before purchase.

The pricing data showed five plan families. Free is listed at $0 and includes 3 Basic Usage units, a 5,000 autocomplete limit, 1,000 advanced model request limit, 10 fast premium model requests, 50 slow premium model requests, and a SOLO agent parallel limit of 2, but SOLO agent, builder, and coder feature flags were not enabled in the extracted Free quota. Lite is listed at $3 monthly, with a $2.25 discounted annualized price also present, and includes 5 Basic Usage units plus SOLO-related feature flags. Pro is listed at $10 monthly, with $7.50 annualized and a $15 period record also present; its extracted quota includes 20 Basic Usage units, 600 fast premium model requests, unlimited slow premium model requests, and a SOLO parallel limit of 10. Pro+ is listed at $30 monthly, $22.50 annualized, and a $45 period record, with 90 Basic Usage units, early access, and a SOLO parallel limit of 15. Ultra is listed at $100 monthly, $75 annualized, and a $150 period record, with 400 Basic Usage units, early access, and a SOLO parallel limit of 20.

Those numbers matter because coding agents can become expensive quickly. A buyer should not compare Trae to Cursor, Windsurf, Aider, Continue, or CodeRabbit by sticker price alone. Convert the plan into a working metric: accepted pull requests per month, successful bug fixes per Basic Usage unit, review minutes saved per task, and failed AI requests that still consume quota. The pricing FAQ explicitly raises Basic Usage, Bonus Usage, different model consumption, running out of usage, Pro trials, failed requests, and refunds, so these are the right questions to ask during evaluation.

Trae is strongest for developers who can provide a clear task and inspect every diff. Good tests include a small bug fix with failing tests, a contained refactor across 3-5 files, a documentation update tied to code changes, or a UI change with screenshot review. Avoid starting with migrations, auth flows, payment logic, data deletion, or production incident work. For the first 20 tasks, track success rate, lines changed, test pass rate, manual cleanup time, and whether the selected plan would have covered the usage.

The upside is speed and focus: Trae can reduce boilerplate and context switching when the repo, tests, and acceptance criteria are clear. The risks are the usual coding-agent risks: hallucinated APIs, overbroad diffs, hidden quota burn, unclear model routing, and privacy questions around repository indexing. Before team rollout, verify private-repo handling, data retention, admin controls, supported IDEs or surfaces, MCP/tool integration, audit logs, and whether failed requests consume Basic or Bonus Usage. Compare it with /tools/coderabbit for code review automation, /tools/qodo for testing quality, /tools/continue-dev for open customization, /tools/aider for terminal-first pairing, and /tools/windsurf for an AI-native IDE.