ToolDriftRE-COUNTED EVERY NIGHT
COUNTED 09:20 UTC4 READS FAILEDRead API
59 models ranked12 boards36 tools tracked, 7 not what they werecaptured 25 Sep 2026against 24 Sep 202629 rose · 23 fell · 51 held · 8 newadoption, not capability
COMPARISONS/CURSOR ALTERNATIVES, ON PUBLISHED PRICES

Cursor alternatives, on published prices

The six tools that answer the same brief as Cursor, on one axis set, with every figure traced to the vendor's own pricing page and the day it was read.

COLUMNS
7
CRITERIA
4
SOURCES
8
VERIFIED
18 September 2026
AI CODING TOOLSEVERY FIGURE READ OFF THAT COMPANY’S OWN PAGE AND DATEDRE-PROBED ON A SCHEDULE
THE VERDICTevery figure in it is a cell in the table below

Four reasons to leave Cursor, and each one picks a different tool. Cost, at team size: Augment Code is "$20 /month flat, no per-seat charge" on Standard for up to 50 seats, so a twenty-developer team pays $20 against $800 for Cursor Teams at $40 / user / mo. Augment Business is the same shape at $100 /month. Cost, at individual size: GitHub Copilot Pro is $ 10 USD per user / month, half of Cursor's $20, and it is already on an invoice your company approved. Control: Zed's Personal plan is "$0 forever" with unlimited use of your own API keys, and Cline charges no subscription at all, "Pay only for AI inference on a usage basis", so on both the only bill is the model bill. Shape of work: Claude Code is a terminal agent rather than an editor, included in the same $20 Pro subscription as the chat product, and it composes with whatever editor you already have. What none of them is, is a cheaper Cursor, every one of these is a different product with a different failure mode. One thing to check before you move: windsurf.com/pricing now redirects to devin.ai/pricing, so a comparison written before that is describing something that no longer exists in that form. ToolDrift sells none of these.

SIDE BY SIDE4 criteria · 7 columns · verified 18 September 2026
COMPARED ONCURSORWINDSURFGITHUB COPILOTCLAUDE CODEZEDCLINEAUGMENT CODE
Free tierwhat the vendor gives you at $0 a month, in its own wordsHobby, Free"No credit card required", limited Agent requests, access to Composer. Cursor serves two pricing layouts at random on the same URL, and the plan name sits next to a tagline in one of them, so this row matches the feature line both layouts carry.cursor.com, read 18 September 2026Free $0⚠ windsurf.com/pricing redirects to devin.ai/pricing, and the plan table it serves is Cognition's Devin, the editor appears on it as Devin Desktop. Read 2026-08-14. The free tier is a "Light quota to code with agents", limited model availability, unlimited Tab completions.windsurf.com, read 18 September 2026$ 0 USD, 2,000 completions per month"Access to Haiku 4.5, GPT-5 mini, and more", Copilot CLI, community support, no credit card.github.com, read 18 September 2026$0, Free for everyoneThe free Claude plan does not include Claude Code; the page lists Claude Code from Pro upwards.claude.com, read 18 September 2026$0 forever (Personal)2,000 accepted edit predictions, and "Unlimited use with your API keys or external agents like Claude Agent, Codex CLI, and more".zed.dev, read 18 September 2026Free, the open-source extension"You only pay for AI inference on a usage basis - no subscriptions, no vendor lock-in."cline.bot, read 18 September 2026not publishedAugment publishes three plans: Standard, Business and Enterprise. None is free and there is no free tier on the page.augmentcode.com, read 18 September 2026
Cheapest paid plan$20 / mo. (Pro)Cursor names the $20 tier Pro. Sampled ten times tonight, the page returned two layouts five times each: one prints Hobby Free, Pro $20 / mo., Pro+ $60 / mo. and Ultra $200 / mo. as four cards, the other prints one Individual card at $20 / mo. covering Pro, Pro+ and Ultra, which is what this row recorded on 14 August 2026. Both agree the cheapest paid price is $20 / mo. Cursor also publishes Start at ₹649/mo, tax inclusive, for developers in India only.cursor.com, read 18 September 2026$20 per month (Pro)Adds "Full model availability", frontier models from OpenAI, Anthropic and Google, and access to cloud agents. Max is $200 per month.windsurf.com, read 18 September 2026$ 10 USD per user / month (Pro)Adds the cloud agent and code review, unlimited completions, and third-party agents, the page names Claude Code and Codex. Pro+ is $ 39 USD and Max is $ 100 USD.github.com, read 18 September 2026$20 if billed monthly (Pro)$17 per month with the annual discount, $200 billed up front. Pro "Includes Claude Code".claude.com, read 18 September 2026$10 per month (Pro)Unlimited edit predictions and "$5 of tokens included", with Zed-hosted models.zed.dev, read 18 September 2026not publishedThere is no paid subscription tier to name. The only plan beside Open Source is Enterprise, priced "Custom".cline.bot, read 18 September 2026$20 /month flat, no per-seat charge (Standard)Augment added a Standard plan below Business after this row was read on 14 August 2026. Standard is $20/month flat with "$20/mo of usage across LLM, Context Engine & compute" included. Business is the same shape at $100/month with $100 of usage.augmentcode.com, read 18 September 2026
Team plan$40 / user / mo. (Teams, Standard seat)"There are two business plans: Teams and Enterprise (custom). Teams offers two seat types: standard ($40/user/mo) and premium ($120/user/mo), where premium adds 5x the standard limits on Agent." Read off Cursor's pricing documentation rather than the pricing page, because one of the two layouts that page serves at random omits the Teams block entirely.cursor.com, read 18 September 2026$80 /month for the team plan plus $40/mo per full dev seatUnlimited team members; each full seat carries its own quota. Enterprise is "Let's talk".windsurf.com, read 18 September 2026not publishedThe plan grid carries a "For businesses" tab, and this page's own text prices only the individual plans, Free, Pro, Pro+ and Max. Nothing readable on it prices Copilot Business or Copilot Enterprise.github.com, read 18 September 2026$25 if billed monthly, per seat (Team)$20 per seat / month billed annually. A Premium seat is $125 monthly or $100 annually and carries "5x more usage than standard seats*".claude.com, read 18 September 2026$30 per seat, per month (Business)Adds org-wide model policies, data governance and unified spend. "Business seats don't bundle a fixed LLM credit allotment."zed.dev, read 18 September 2026not publishedEnterprise is the team offering and its price is "Custom", the page's call to action is "Contact Sales".cline.bot, read 18 September 2026$20/month flat, up to 50 seats included (Standard)"One flat price for your whole team, up to 50 seats with $20 of usage included every month." Business is the same shape at $100/month with $100 of usage. Enterprise is Custom with unlimited users. Usage and top-ups are pooled across the team, with no per-seat charge.augmentcode.com, read 18 September 2026
Past the included usagewhat the vendor says happens when the plan's allowance runs out, the line that decides the real bill, and the one no headline price includesOn-demand usage, billed in arrears"Every plan includes a set amount of model usage. On-demand usage allows you to continue using models after your included amount is consumed, billed in arrears."cursor.com, read 18 September 2026Buy extra usage at API pricing"Each paid plan comes with a usage allowance that refreshes automatically on a daily and weekly basis... If you do go beyond your included usage on a paid plan, you can purchase extra usage which is consumed at API pricing."windsurf.com, read 18 September 2026GitHub AI Credits, $15 monthly total credits for ProCredits are consumed by chat, agent mode, code review and the cloud agent. Pro+ carries $70 of monthly credits and Max $200.github.com, read 18 September 2026not publishedThe plans are described in multiples rather than rates, "More usage*", "Choose 5x or 20x more usage than Pro*", and the asterisk points at "Usage limits apply." No per-unit price for exceeding a plan is printed on this page.claude.com, read 18 September 2026Usage-based billing beyond the $5 of tokens includedOr bring your own keys, "in which case spend goes directly to your provider".zed.dev, read 18 September 2026You pay for inference, at cost, or on your own key"You can bring your own API keys from providers like OpenAI, Anthropic, Google, and others. This gives you complete control over your costs and usage."cline.bot, read 18 September 2026Top-ups, pay as you goPast the usage included in the plan, $20 on Standard or $100 on Business, you buy top-ups rather than moving to a higher tier.augmentcode.com, read 18 September 2026
WHEN TO PICK EACHthe case for each column, in its favour

Pick Cursor when

You want the editor everyone else on the team already knows, and you want the agent and the editor to be one product rather than a plugin. Cursor is the default answer for a reason: it is the most widely used of these, its $20 tier is the same price as the cheapest agent subscription here, and its Teams tier at $40 / user / mo. is the only one on this table that bundles SSO, pooled billing and agentic code review without a sales call.

Pick Windsurf when

You want the agent to keep working after you close the laptop. Cognition sells Windsurf and Devin as one thing now, so the $20 tier buys cloud agents rather than only a local editor, and the Max tier at $200 per month is the only published price in this table aimed squarely at someone running long autonomous jobs rather than pair-programming.

Pick GitHub Copilot when

The code already lives in GitHub and you would rather add a line to an invoice you are already paying than start a new vendor relationship. Copilot is the cheapest paid tier in this table at $ 10 USD per user / month, it is the only one whose agent runs inside pull requests and code review natively, and it is the one your procurement department has already approved.

Pick Claude Code when

The work is agentic rather than autocomplete, long refactors, migrations, a task you describe once and come back to. Claude Code is a terminal agent rather than an editor, so it composes with whatever editor you already use instead of replacing it, and it is included in the same $20 Pro subscription as the chat product rather than being a separate bill.

Pick Zed when

You already pay for model access somewhere else and you resent paying twice. Zed's Personal plan is "$0 forever" with "Unlimited use with your API keys or external agents like Claude Agent, Codex CLI, and more", it is the only editor in this table whose free tier is a complete product rather than a trial, and it is open source, so bringing your own key is a supported path rather than a workaround.

Pick Cline when

You want the agent's costs to be your model bill and nothing else, and you want to see every token it spends. Cline is open source, charges no subscription at all, "Pay only for AI inference on a usage basis - no subscriptions, no vendor lock-in", and it runs against Anthropic, OpenAI, Gemini, OpenRouter, Bedrock and Vertex, so nobody between you and the model is taking a margin you cannot see.

Pick Augment Code when

The codebase is large enough that context is the whole problem, and the team is big enough that per-seat pricing has stopped being rational. Augment is the only entry in this table priced as "$20 /month flat, no per-seat charge" on Standard for up to 50 seats, so a twenty-developer team pays $20 against $800 for Cursor Teams, and the retrieval engine is what it sells rather than the editor.

THE QUESTIONS4 answered from the table above

What is the cheapest alternative to Cursor?

Free ones, and they are real products rather than trials: Zed's Personal plan is "$0 forever" with "Unlimited use with your API keys or external agents like Claude Agent, Codex CLI, and more", and Cline is free for individual developers with no subscription at all. Of the paid tiers, GitHub Copilot Pro at $ 10 USD per user / month is the cheapest, half Cursor's $20 / mo. On both of the free ones you still pay for model inference, the saving is the subscription, not the tokens.

Which Cursor alternative is best for a team?

Augment Code on price, GitHub Copilot on procurement. Augment Standard is "$20 /month flat, no per-seat charge" with "up to 50 seats included", the only non-per-seat price in this comparison, against Cursor Teams at $40 / user / mo. for a Standard seat. That is cheaper from the first developer, and Augment Business is the same shape at $100 /month. Copilot is cheaper per head and already inside a GitHub bill, but its own plans page carries only the individual tiers; the business ones sit behind a "For businesses" tab and are not on the page's text at all.

Do any of these let me use my own API key?

Zed and Cline, explicitly. Zed's free plan says "Unlimited use with your API keys or external agents", and on Business "you can use Zed-hosted AI at standard usage rates, or bring your own API keys from Anthropic, OpenAI, Google, and others - in which case spend goes directly to your provider". Cline lists "Purchase and Model Inference at Cost or BYOK" and names Anthropic, Gemini, OpenAI, OpenRouter, AWS Bedrock, GCP Vertex and others. Cursor, Windsurf, Copilot and Claude Code sell you bundled usage instead.

Is Windsurf still a Cursor alternative?

It is, but not the one people mean. windsurf.com/pricing redirects to devin.ai/pricing and the plan table there is Cognition's Devin, with the editor listed as "Devin Desktop". So switching to Windsurf today means buying into Devin's plan ladder, Free, Pro at $20 per month, Max at $200 per month, and what the money buys is cloud agents rather than a better editor. Read on 2026-08-14. This is the exact class of change ToolDrift exists to catch: the comparison articles still in circulation were written before it.

WHERE EACH FIGURE CAME FROM8 pages, each on the company’s own domain
COLUMNPAGEWHAT IT SUPPLIEDREAD
Cursorcursor.comFree tier, Cheapest paid plan, Past the included usage18 September 2026
Cursorcursor.comTeam plan18 September 2026
Windsurfwindsurf.comFree tier, Cheapest paid plan, Team plan, Past the included usage18 September 2026
GitHub Copilotgithub.comFree tier, Cheapest paid plan, Team plan, Past the included usage18 September 2026
Claude Codeclaude.comFree tier, Cheapest paid plan, Team plan, Past the included usage18 September 2026
Zedzed.devFree tier, Cheapest paid plan, Team plan, Past the included usage18 September 2026
Clinecline.botFree tier, Cheapest paid plan, Team plan, Past the included usage18 September 2026
Augment Codeaugmentcode.comFree tier, Cheapest paid plan, Team plan, Past the included usage18 September 2026
  • Cursorcursor.comFree tier, Cheapest paid plan, Past the included usageREAD 18 SEPTEMBER 2026
  • Cursorcursor.comTeam planREAD 18 SEPTEMBER 2026
  • Windsurfwindsurf.comFree tier, Cheapest paid plan, Team plan, Past the included usageREAD 18 SEPTEMBER 2026
  • GitHub Copilotgithub.comFree tier, Cheapest paid plan, Team plan, Past the included usageREAD 18 SEPTEMBER 2026
  • Claude Codeclaude.comFree tier, Cheapest paid plan, Team plan, Past the included usageREAD 18 SEPTEMBER 2026
  • Zedzed.devFree tier, Cheapest paid plan, Team plan, Past the included usageREAD 18 SEPTEMBER 2026
  • Clinecline.botFree tier, Cheapest paid plan, Team plan, Past the included usageREAD 18 SEPTEMBER 2026
  • Augment Codeaugmentcode.comFree tier, Cheapest paid plan, Team plan, Past the included usageREAD 18 SEPTEMBER 2026

Every cell above names the exact page it was read off, the day it was read, and a literal string that must still be on that page for the figure to hold. A job re-fetches each one on a schedule and fails when the string is gone, which is the difference between a table that is checked and one that was written once.