Updated on
Trae: reviews and analysis
ByteDance's AI IDE: frontier models at rock-bottom prices.
Affiliate link · no extra cost to you
Our verdict
Trae competes on price by a wide margin: access to frontier models for a fraction of what rivals charge, on a VS Code-derived IDE with a built-in agent. The math is so good that the question shifts to the fine print: it is a ByteDance product, and each team must decide whether its data policy fits that provenance. On capability, it delivers without leading.
Best for: Price-sensitive developers untroubled by the product's provenance.
What changed this week
Sweep of
No substantive change from the previous sweep: what we checked this week confirms what was already there.
- no change
- no change
We have tracked it since July 25, 2026: 3 sweeps on record. We publish the latest change here; the full series is not published.
What the internet says
It remains the cheapest agent editor in the catalogue: the paid step is still $3 a month with a free plan below it, and this week brought no product or price change that would move the score. The underlying problem is unchanged: its documentation still says your chats and the code inside them feed model training unless you turn on privacy mode yourself, so it stays off-limits for client code or regulated work.
What the web repeats in favour
- The cheapest paid step in the catalogue, $3 a month, with a free plan below it to try it out
- It is a VS Code derivative, so the environment feels familiar from minute one
- There is a 7-day free trial of the middle plan for new sign-ups, though only if you pay by credit card, you have to claim it within 48 hours and it auto-renews into a paid plan unless you cancel
- Development moves fast and ships real agent plumbing: hooks, subagents, slash commands, nested rules and MCP, with a version out on 30 July 2026
What the web repeats against
- For client code, sensitive intellectual property or regulated work the repeated advice is not to use it without security approval, and the vendor's own documentation backs that up: your chats and the code snippets inside them feed model training unless you turn privacy mode on yourself, and it stops applying once you log out
- Telemetry is the central objection: a widely discussed analysis documented collection of file paths and hardware identifiers with the setting turned off
- Consumption billing makes spend hard to predict, and the bonus allowance moves without notice: its official repo has complaints that the high plan's extra was cut
- Opting out of auto-renewal costs you: a single month of the entry plan costs half as much again as the recurring one, and monthly subscriptions get no refund once you have used anything
- What you read about Trae in Chinese is not what you buy here: the domestic version charges in another currency and by credits, and its entry plan does not even cover the same product
Sweep sources: Official pricing · Docs · Docs · GitHub · Communities
Pros / Cons
Pros
- The category's lowest price by far
- Familiar VS Code-derived IDE
- Free plan with broad autocomplete
Cons
- ByteDance's data policy stops many teams
- Capability behind the leaders
TLDR: Trae competes on price by a wide margin: access to frontier models for a fraction of what rivals charge, on a VS Code-derived IDE with a built-in agent. Broad free plan, Lite at $3 and Pro at $10, against everyone else’s $20. The math is so good the question shifts to the fine print: it is a ByteDance product, and each team decides whether its data policy fits.
What Trae is and how it works
Trae is ByteDance’s move into AI IDEs, executed with the strategy the company has mastered: a competent product at a market-breaking price. On a VS Code-derived base (extensions and habits migrate on their own), it offers what defines the category: an agent for tasks, autocomplete, project-context chat and access to frontier models from several providers.
The pricing structure is the headline: a free plan with broad autocomplete and access to serious models, a Lite tier at $3 and Pro at $10, half or less of what Cursor or Devin Desktop charge for access to comparable model families.
The question accompanying the product is its parent company’s, and it is not rhetorical: for company data, the provider’s processing policy and jurisdiction are part of the evaluation, and they weigh more here than in other cases.
What it is like day to day
Daily experience is that of a competent AI IDE: the agent solves medium tasks, autocomplete delivers and the integration is more polished than the price suggests. It leads in nothing and delivers in almost everything, which at $10 is exactly the offer.
The honest comparison with the leaders: Cursor’s agent keeps the edge in autonomy and finesse, and Trae’s edge is the bill. For students, personal projects and tight budgets, that $10-a-month difference is real. For the professional living off the agent, the capability gap is usually worth more.
Daily life also includes the informed decision: reading the data policy, deciding which projects pass through and which do not. That is an exercise worth doing with any provider, one this product’s provenance makes explicit.
Pricing and plans
A free plan with broad autocomplete and model access, Lite at $3 a month and Pro at $10 with extended limits and the frontier model families. It is the most aggressive pricing structure among AI IDEs, by a distance.
The math against the $20 rivals is immediate: half the price for an experience that delivers. The complete math includes the data variable, which has no price on the table and every team weighs differently.
Who it is for (and who it is not for)
Trae is for price-sensitive developers untroubled by the provenance: students, personal projects, markets where $20 a month weighs. Also as a low-cost secondary IDE alongside other tools.
It is not for teams whose data policies exclude the provider, nor for anyone after the market’s best agent, which still lives in other houses.
Trae alternatives
Cursor and Devin Desktop are the category leaders at double the price. GitHub Copilot competes on entry price ($10) with GitHub integration as its card. Kimi CLI plays the same cost card from the terminal. They are in Trae alternatives and the ranking in the best AI coding tools.
Frequently Asked Questions
How can it cost half of Cursor?
ByteDance’s entry strategy: aggressive pricing to win share, subsidizing model costs. How long that structure lasts is each user’s bet, and while it lasts, the math is what it is.
Does the agent compete with Cursor’s?
It delivers on medium tasks without reaching the leader’s autonomy or finesse. The offer is not being the best, it is delivering 85% for half the price.
Is it safe for company code?
That is the right question and its answer belongs to each team: read the current data policy, weigh the provider’s jurisdiction and decide which projects pass through. For personal and open source code, most find no objection.
Trae or Copilot at $10?
Same price, different offers: Copilot gives the GitHub integration and its ecosystem, Trae gives the full IDE with access to more model families. The data variable tips the scale depending on each context.