Every price read from cursor.com and opencode.ai
OpenCode vs Cursor: OpenCode Cursor Pricing and Cursor vs OpenCode Cost per Month
Every comparison of these two opens with "free versus $20 a month". That gets the money backwards. Both end in a token bill, and the only per-token markup in this pairing belongs to the paid tool, on your own API key.
Flat $29/mo billed yearly · No API key · No usage meter
Pick a task
Plan
- planning
Files changed
Test run
Pull request
You review and merge. AgentCode never merges on its own.
The short answer
Cursor costs $20 a month for Pro, $60 for Pro Plus and $200 for Ultra, but that fee buys the editor, unlimited Tab completions, Bugbot and a pool of Cursor's own models. Third-party models such as Claude and GPT are charged on top at their API price. OpenCode costs nothing to install and its Zen gateway sells tokens at cost, so you pay the same model rates without the subscription. Cursor's own docs put a daily agent user at $60 to $100 a month in total usage, which is the number to compare, not the $20.
OpenCode vs Cursor at a glance
One is an editor you buy and one is an agent you install. They overlap on the part most people are actually paying for, which is a model doing the work, and they differ on everything wrapped around it.
Swipe to see more
| OpenCode | Cursor | |
|---|---|---|
| What the tool itself costs | Nothing. MIT-licensed terminal agent and desktop app. | $20 Pro, $60 Pro Plus, $200 Ultra per month. |
| What that fee includes | No fee, so nothing is bundled. | Editor, unlimited Tab, Bugbot, Cloud Agents and the Cursor Models pool. |
| How third-party models are billed | At the provider's list rate, on your key or through Zen at cost. | At the model's API price out of the Other Models pool, then on demand at the same rates. |
| Markup on your own API key | None. Enterprise adds no token charge on your own gateway. | $0.25 per million tokens on Teams and Enterprise, BYOK included. |
| Where it runs | Terminal, desktop app, any editor you already use. | Its own VS Code-derived editor, plus a CLI and cloud agents. |
| Models | 75+ providers including local models through Ollama. | A curated list: Claude, GPT, Gemini, Grok, Composer. No local models. |
| Cheapest paid rung | $10 a month for OpenCode Go, open models only. | $20 Pro. The Rs 649 Start plan is sold only in India. |
| Team pricing | Enterprise per seat on a quote. Zen workspaces free during beta. | Teams at $40 or $120 per user a month, Enterprise on a quote. |
| Spend control | Workspace and per-member monthly limits set by an admin. | Included pool, then on-demand usage you opt into. |
| Who owns the output | You. The docs state no licensing restrictions or ownership claims. | You, under Cursor's terms. |
The markup runs the other way
Here is the part that almost every article about this pairing misses, and it only shows up once a team is involved. Cursor's documentation describes a Cursor Token Rate of $0.25 per million tokens that applies to third-party model requests on Teams and Enterprise plans. Its own wording is that the rate applies to included usage, to on-demand usage, and to bring-your-own-key usage. Cursor's first-party models, Grok and Composer, are exempt from it.
OpenCode's enterprise documentation says the opposite in one sentence: if you have your own LLM gateway, it does not charge for tokens used. Its Zen gateway is described as selling at cost, with the only markup covering card processing, passed through at 4.4% plus $0.30 per transaction.
The free tool
No token margin at all
On your own key you pay the provider. Through Zen you pay list plus card fees. Through your own gateway on Enterprise, OpenCode states it charges nothing for tokens.
The paid tool
$0.25 per million, on your own key
On Teams and Enterprise the Cursor Token Rate sits on top of the model's API price, and Cursor's docs say BYOK usage is included in it.
What it means
The $20 is not the decision
For one developer the subscription dominates. For a team on a shared key, the per-token rate does, and it is the paid tool that carries it.
What Cursor's $20 actually buys
Cursor runs two separate usage pools that reset with your billing cycle. The Cursor Models pool covers its own models, Grok 4.6, Grok 4.5 and Composer 2.5, and gets what Cursor calls significantly more included usage. The Other Models pool covers everything from Anthropic, OpenAI and Google, and those are charged at the model's API price. So the subscription is buying product and a subsidised first-party pool, not a flat allowance of frontier tokens.
Swipe to see more
| Plan | Price | Included | What to know |
|---|---|---|---|
| Pro | $20/mo | Both usage pools | Unlimited Tab completions, extended agent limits on all models, Bugbot and Cloud Agents. The Other Models pool covers third-party usage until it runs out, after which you add on-demand usage at the same API rates. |
| Pro Plus | $60/mo | A larger share of both pools | The rung most daily agent users land on, because Cursor's own guidance puts that workload at $60 to $100 a month of total usage. Same feature set as Pro, more included usage. |
| Ultra | $200/mo | The largest individual allowance | Aimed at the profile Cursor calls power users, running multiple agents and automations, which its docs put at $200 or more a month of usage. |
| Teams, standard seat | $40/user/mo | Team pools plus admin | Adds centralized billing, a team marketplace for rules and plugins, Bugbot code review, shared team context for cloud agents, usage analytics, privacy-mode enforcement and SAML or OIDC SSO. The Cursor Token Rate starts here. |
| Teams, premium seat | $120/user/mo | 5x standard limits on Agent | Three times the price for five times the agent allowance, so unlike most of this market it is a genuine volume discount rather than a flat multiple. |
| Start | Rs 649/mo | Cursor Models pool only | Sold only to developers in India, billed in INR. No Other Models pool, no on-demand usage, no Bugbot, no Auto and no SDK. US comparisons that quote it as Cursor's budget tier are quoting a price a US buyer cannot purchase. |
One more thing worth knowing before you plan work around very large contexts: Max Mode, which extends a model's context window and bills at the API rate plus 20%, is now documented as available on legacy request-based plans only. A new subscriber cannot buy it. We keep the wider Cursor price list current on the Cursor pricing page.
What OpenCode costs once you use it
The harness is MIT-licensed and free, so the question is only which meter you attach to it. There are five, and they differ enormously in what they cover.
Swipe to see more
| Route | Price | Meter | What to know |
|---|---|---|---|
| Bring your own API key | Provider list rates | Billed by the provider | Anthropic, OpenAI, Google, Bedrock, Vertex, OpenRouter, Groq, Azure, Together and any OpenAI-compatible endpoint, plus local models through Ollama. OpenCode adds nothing to the rate. |
| OpenCode Zen | Pay per request, at cost | Prepaid balance, auto-reload $20 | One key for a tested list of Claude, GPT, Gemini and open models. OpenCode states it sells at cost and passes card fees through at 4.4% plus $0.30 per transaction. Models are hosted in the US on zero-retention providers, with exceptions it lists by name. |
| OpenCode Go | $10 a month | 5-hour, weekly and monthly limits | Open coding models such as Kimi, GLM, Qwen, DeepSeek and MiniMax, plus Grok 4.6 and GPT 5.6 Luna. No Claude and no Gemini. Only one member per workspace can subscribe, and the docs say it is designed primarily for international users. |
| A subscription you already hold | Whatever that plan costs | That plan's own limits | OpenCode signs in to ChatGPT Plus or Pro, GitHub Copilot and GitLab Duo. Claude Pro and Max are the notable exception, because Anthropic prohibits third-party tools from using them. |
| OpenCode Enterprise | Per seat, custom quote | Your own gateway | Central config, SSO integration and a lock that routes every request to your internal AI gateway. If you bring that gateway, OpenCode says it does not charge for tokens used. |
The correction worth making loudly, because the comparison articles all get it wrong: OpenCode Go at $10 is not the budget version of Cursor Pro. It carries open models only, no Claude and no Gemini, just one member per workspace may subscribe, and OpenCode's own documentation says it is designed primarily for international users. For a US team it is a model experiment, not a plan. The same $10 tier is examined against Anthropic's plans on OpenCode against Claude Code.
Cursor vs OpenCode cost per month, by how you actually work
Cursor publishes something most vendors will not: its own estimate of what customers spend. Daily Tab users typically stay within included usage. Daily agent users typically run $60 to $100 a month in total usage. Power users running multiple agents and automations often exceed $200. Those are the numbers to compare against, not the plan price. The OpenCode column below is the same model usage on a provider key, plus the card fee OpenCode passes through; the arithmetic is ours.
Swipe to see more
| How you work | Cursor | OpenCode | Which wins |
|---|---|---|---|
| Tab completions and occasional chat | $20 | Roughly $5 to $15 of tokens | Cursor. Tab is unlimited inside the $20 and OpenCode has no equivalent, so you would be paying tokens for less product. |
| Agent a few times a week | $20, usually inside the pool | Roughly $20 to $40 of tokens | Cursor, narrowly. Light agent use is what the included pool is sized for. |
| Agent every working day | $60 to $100 of total usage, by Cursor's own figure | The same tokens, plus about 4.4% in card fees | Roughly level on tokens. Cursor is worth the difference only if you use Tab, Bugbot and Cloud Agents enough to justify the subscription on top. |
| Multiple agents and automations | $200+ of usage, by Cursor's own figure | The same tokens at list, or open models far below it | OpenCode, if you will accept open models. That is where the gap becomes real money rather than a rounding error. |
| A ten-person team on its own API key | $400 to $1,200 a month in seats, plus $0.25 per million tokens on top of the key | Enterprise seats, and no token charge on your own gateway | This is the one case where the open-source tool is unambiguously cheaper on tokens, because Cursor's Token Rate applies to BYOK usage and OpenCode's does not exist. |
Is OpenCode better than Cursor for your situation?
Model quality is close enough that it rarely decides this, because both can point at the same frontier models. What decides it is whether you are buying a product or buying control.
You live in Tab completions
Cursor Pro
Unlimited Tab is inside the $20 and there is no OpenCode equivalent. This is the clearest case where the subscription is buying you something the free tool does not have.
You drive an agent all day
Compare on usage, not on $20
Cursor puts this profile at $60 to $100 a month of total usage. The subscription is a floor. Run the same model through OpenCode on your key and the token half of the bill is close to identical.
Your team brings its own API key or gateway
OpenCode
Cursor applies a $0.25 per million token rate on Teams and Enterprise that covers BYOK usage too. OpenCode Enterprise says it charges nothing for tokens on your own gateway. On a large key that difference compounds every month.
You want the smallest possible monthly number
OpenCode Go at $10, with a caveat
Go holds no Claude and no Gemini, only one member per workspace can subscribe, and OpenCode says it is designed primarily for international users. Test latency and output quality on real tickets before a US team commits to it.
You need local or self-hosted models
OpenCode
Cursor's model menu is curated and does not include local models. OpenCode talks to Ollama and any OpenAI-compatible endpoint, which is often the deciding factor when legal will not approve an external provider.
You want finished pull requests, not a pair session
A repo agent on a flat bill
Both of these sell time at the keyboard and both end in a meter. If the job is a described task returned as a tested pull request for a person to review, that is a different purchase with a different price shape.
OpenCode Enterprise vs Cursor Teams
Cursor Teams sells two seat types, $40 standard and $120 premium, where the premium seat carries five times the standard agent limits. Three times the price for five times the usage is a genuine volume discount, which is rarer in this market than it should be. The seat also brings centralized billing, a team marketplace for internal rules and plugins, Bugbot code review, shared team context for cloud agents, usage analytics, privacy-mode enforcement and SAML or OIDC single sign-on. Cursor recommends Enterprise instead when you need pooled usage, invoicing, SCIM or priority support.
OpenCode Enterprise is priced per seat on a quote and sells the control layer rather than the model access: a single central config for the organization, SSO integration, and a lock that sends every request through your internal AI gateway with all other providers disabled. OpenCode stores no code or context data, and the one exception it names is the optional share feature, which you can turn off in a config file. Zen team workspaces, with admin and member roles and per-member monthly spend limits, are free while the beta runs.
The practical difference for a finance team is predictability. A Cursor seat is a fixed line that then grows with on-demand usage and the token rate. An OpenCode seat is a fixed line plus whatever your gateway bills, which you already see. Neither gives you a hard ceiling by default, which is the same trade we picked apart for terminal agents on Cline against Claude Code.
Where we fit, and where both tools beat us
The losses first, because they are real. Cursor is a better place to write code by hand than anything we make: Tab completions, inline edits and a design mode are worth the subscription on their own if that is how you work. OpenCode's model freedom, including local models through Ollama, is something we do not offer and cannot match. Both are strong interactive tools for exploratory work where you do not yet know what you want built.
AgentCode is narrower on purpose. You connect a GitHub or GitLab repository, describe a task, and the agent plans it, edits the code, runs your tests and opens a pull request for a person to review. It never merges on its own and never trains on your code. There is no API key to rotate, no provider to choose, no usage pool to watch and no invoice that moves with how busy the sprint was. It is $29 a month billed yearly.
If you are reading a comparison like this because the bill is the problem rather than the tool, how the agent runs a task end to end shows what a flat price buys instead, and our plans list exactly what each tier includes.
OpenCode vs Cursor: the questions people search
Is OpenCode better than Cursor?
Neither is better outright; they sell different things. Cursor sells an editor with unlimited Tab completions, Bugbot code review and cloud agents for a monthly fee. OpenCode sells nothing and gives you a terminal agent that talks to any provider, including local models. Pick Cursor for the product around the model, OpenCode for control of the model and the bill.
Is OpenCode really free?
The software is free and MIT-licensed, but every task calls a model and that is where the bill starts. You pay your provider directly, top up an OpenCode Zen balance, subscribe to Go at $10 a month, or reuse a ChatGPT or Copilot plan you already hold. What OpenCode does not do is add a margin on the tokens.
How much does Cursor actually cost per month?
Pro is $20, Pro Plus $60 and Ultra $200. Cursor's own documentation then gives spend bands that matter more than the plan price: daily Tab users usually stay within included usage, daily agent users typically run $60 to $100 a month in total usage, and power users running multiple agents often exceed $200.
OpenCode Go vs Cursor Pro: which is cheaper?
Go is $10 and Cursor Pro is $20, but they are not the same purchase. Go covers open models only, with no Claude and no Gemini, and OpenCode allows just one member per workspace to subscribe. Cursor Pro includes unlimited Tab, Bugbot, Cloud Agents and access to frontier models. Go is cheaper; it is not a Cursor replacement.
Does Cursor charge extra if I bring my own API key?
On Teams and Enterprise plans, yes. Cursor applies a Cursor Token Rate of $0.25 per million tokens to third-party model requests, and its docs state this applies to included usage, on-demand usage and bring-your-own-key usage. Cursor's own first-party models, Grok and Composer, are exempt. OpenCode Enterprise charges nothing for tokens on your own gateway.
Is OpenCode Zen cheaper than Cursor?
On tokens, marginally. OpenCode states that Zen sells at cost and that its only markup covers processing fees, which it passes through at 4.4% plus $0.30 per transaction. Cursor charges third-party models at the model's API price out of your included pool, so the per-token rates are similar. The difference is the subscription sitting on top of Cursor and the token rate on its team plans.
OpenCode vs Cursor CLI: what is the difference?
Both put an agent in the terminal. Cursor's CLI is tied to your Cursor account and spends your plan's usage pools, so it needs a subscription. OpenCode's is the whole product, needs no account, and spends whichever provider credential you configure. If you are already paying Cursor, the CLI is included; if you are not, OpenCode gets you a terminal agent without starting a subscription.
Can I use Cursor and OpenCode together?
Yes, and plenty of developers do. Cursor is an editor and OpenCode runs in the terminal alongside whatever editor you use, so keeping Cursor for Tab completions and inline edits while running longer agent tasks through OpenCode on your own key is a common arrangement. You pay the Cursor subscription and the tokens separately.
Does OpenCode work with Claude like Cursor does?
It reaches the same Claude models at the same API rates, through your Anthropic key or through Zen, but it cannot use a Claude Pro or Max subscription because Anthropic prohibits third-party tools from doing that. Cursor is in the same position: Claude inside Cursor is billed out of your usage pool, not out of a Claude plan.
What is the Cursor Models pool?
It is Cursor's first-party usage pool, covering Grok 4.6, Grok 4.5 and Composer 2.5, and Cursor gives it significantly more included usage than the pool for third-party models. On Teams and Enterprise, Cursor Router picks which of these to use for each Auto request. These models are also exempt from the Cursor Token Rate.
Does Cursor still have Max Mode?
Only on legacy request-based plans. Cursor's documentation now lists Max Mode, which extends a model's context window beyond the default and bills at the model API rate plus 20%, as available on those older plans alone. If you are signing up today you cannot buy it, which is worth knowing before you plan work around very large contexts.
Can a team control spending on OpenCode?
Yes. A Zen workspace admin can set a monthly usage limit for the whole workspace and for each individual member, enable or disable specific models, and manage API keys and billing. Workspaces are free for teams during the beta. The caveat worth planning around is auto-reload, which can push you past a monthly limit if the balance drops below $5.
Keep researching
The Cursor side in more depth: Cursor pricing across every plan, Cursor against Claude Code, Codex against Cursor and what Cursor costs per month in practice.
The open-source harnesses against the vendor tools: OpenCode against Claude Code, OpenCode against Codex on a ChatGPT plan, Cline against Claude Code, Cline against Cursor and Cline against Codex.
A repo agent with no usage pool to watch
Connect a GitHub or GitLab repository, describe a task, and review the pull request it opens. Flat $29/mo billed yearly, no API key, no token rate, and nothing merges without you.