Pricing read from opencode.ai and Anthropic's own pages
OpenCode vs Claude Code: Claude Code vs OpenCode Pricing, OpenCode Claude Subscription and Plans
OpenCode is free to install and Claude Code comes with a $20 to $200 plan, so the free one looks cheaper. It is not, for Claude. Your Claude plan does not work inside OpenCode, which means every Claude token there is billed at list rate. The saving is real only when you switch models.
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
OpenCode is free, open-source software, but Claude inside it is always billed per token, because Anthropic prohibits using a Claude Pro or Max plan in third-party tools. Claude Code comes with those plans at $20, $100 or $200 a month. So for someone who codes with Claude most days, Claude Code is the cheaper way to buy Claude. OpenCode wins when you want other models: its $10 Go plan, or a ChatGPT or Copilot subscription you already pay for.
OpenCode vs Claude Code at a glance
Both are terminal-first coding agents that read your repository, edit files and run commands. The difference that decides the bill is which subscriptions each one is allowed to spend. OpenCode takes almost every major one except Claude's. Claude Code takes only Claude's.
Swipe to see more
| OpenCode | Claude Code | |
|---|---|---|
| What you pay for the tool | Nothing. MIT-licensed, runs in the terminal and a desktop app. | Included with Claude Pro, Max, Team and Enterprise. |
| Can it use a Claude Pro or Max plan? | No. OpenCode removed the plugins in 1.3.0 and says Anthropic explicitly prohibits it. | Yes, that is what the plan is for. |
| Subscriptions it accepts | ChatGPT Plus and Pro, GitHub Copilot, GitLab Duo, and its own $10 Go plan | Claude Pro, Max, Team and Enterprise |
| Claude Sonnet 5 per million tokens | $2 in, $10 out, on your key or through Zen | $2 in, $10 out on a Console key, or inside your plan quota |
| Cheapest paid way in | $10 a month OpenCode Go, no Claude models | $17 a month, Claude Pro billed annually |
| Models | 75+ providers: Claude, GPT, Gemini, Grok, Kimi, GLM, Qwen, DeepSeek, local | Claude, or another model through an LLM gateway |
| Spend ceiling | Go limits, or a monthly limit you set in a Zen workspace | Your plan cap on a subscription, none on an API key |
| Team plan | Enterprise per seat, no token charge on your own gateway | Team for 2 to 150 seats at $25 or $125 a month; Enterprise bills usage at API rates |
Can I use OpenCode with a Claude subscription?
No, not in any current version, and this one fact settles most of the price comparison. OpenCode's own provider documentation says there are plugins that let you use Claude Pro or Max with OpenCode, that "Anthropic explicitly prohibits this", and that OpenCode stopped bundling those plugins as of version 1.3.0.
What Anthropic says
Plan sign-in is for Anthropic's own apps
Its Claude Code legal page says subscription OAuth is for ordinary use of Claude Code and native Anthropic applications, and that third-party developers may not route requests through Free, Pro or Max credentials.
What OpenCode did
Removed the Claude Pro/Max plugins in 1.3.0
The sanctioned ways to reach Claude from OpenCode are now an Anthropic API key, a cloud provider such as Bedrock or Vertex, or OpenCode Zen. Every one of them is metered per token.
What that costs you
No cap on Claude spend
A Claude plan stops you at a limit and the bill stays flat. Claude in OpenCode keeps going and so does the invoice, unless you set a monthly limit in a Zen workspace yourself.
OpenCode makes the contrast explicit in the same paragraph: "Other companies support freedom of choice with developer tooling", and it lists ChatGPT Plus, GitHub Copilot and GitLab Duo as subscriptions that work in OpenCode with zero setup. If you already pay OpenAI, GitHub or GitLab, OpenCode is a harness you get for free on money you are already spending. If you pay Anthropic, it is not.
OpenCode pricing: six ways to pay for a free tool
The software costs nothing and OpenCode does not store your code. The bill comes from the model, and OpenCode offers several ways to pay for one. They behave very differently at the end of a month, so it is worth knowing which one people mean when they say "OpenCode subscription".
Bring your own API key
Provider list rates
Per token, billed by the provider
Anthropic, OpenAI, Google, Bedrock, Vertex, OpenRouter and dozens more. OpenCode takes nothing, so Claude Sonnet 5 is $2 per million input tokens and $10 per million output, straight from Anthropic.
OpenCode Zen
Pay as you go, at list rate
Prepaid balance, auto-reload $20
One key for a tested list of Claude, GPT, Gemini and open models. OpenCode says it sells at cost and passes card fees through at 4.4% plus $0.30 per transaction. Models are hosted in the US.
OpenCode Go
$10 a month
5-hour, weekly and monthly limits
27 models, mostly open ones such as Kimi, GLM, Qwen, DeepSeek and MiniMax, plus GPT 5.6 Luna and Grok 4.6. Each model gets a monthly dollar allowance of $15, $30 or $60. No Claude models.
A subscription you already have
Whatever that plan costs
That plan's own limits
OpenCode signs in to ChatGPT Plus or Pro, GitHub Copilot and GitLab Duo directly. Some Copilot models need Pro+. Claude plans are the one major subscription missing from this list.
OpenCode Black
Enrollment paused
Not currently sold
OpenCode's top plan for access to Claude, GPT and Gemini. Its page says enrollment is temporarily paused, so you cannot buy it today.
OpenCode Enterprise
Per seat, custom quote
Your own gateway
Central config, SSO and a lock to your internal AI gateway. OpenCode says it does not charge for tokens when you bring your own gateway.
Zen deserves one extra line of arithmetic. Its Claude rates match Anthropic's list exactly, but card fees of 4.4% plus $0.30 are passed through on each top-up. On the default $20 auto-reload that is $1.18, or just under 6% on top of list. Buying Claude through Zen is convenient, never cheaper.
OpenCode Go: what $10 a month actually buys
Go is the plan most people mean by an OpenCode subscription. It is $10 a month for 27 models, and it is not unlimited: each model has its own monthly allowance measured in dollars, and you can spend at most 20% of it in any five-hour window and 50% in a week. Cheap models stretch a long way, expensive ones do not.
Swipe to see more
| Model on Go | Monthly usage included | Est. requests per 5 hours | Rate, in / out per million |
|---|---|---|---|
| Kimi K3 | $15 | 110 | $3.00 / $15.00 |
| GPT 5.6 Luna | $15 | 2,050 | $0.20 / $1.20 |
| Grok 4.6 | $15 | Not published | $2.00 / $6.00 |
| Qwen3.8 Max | $15 | 160 | $2.00 / $6.00 |
| GLM-5.3 | $15 | 220 | $1.40 / $4.40 |
| Kimi K2.7 Code | $60 | 1,350 | $0.95 / $4.00 |
| MiniMax M3 | $60 | 3,200 | $0.30 / $1.20 |
| Qwen3.7 Plus | $60 | 4,300 | $0.40 / $1.60 |
| DeepSeek V4 Flash | $30 | 13,000 | $0.30 / $1.20 peak |
Two things the comparison articles miss. First, Go is no longer only open models: GPT 5.6 Luna and Grok 4.6 are on the list, each with a $15 allowance. Second, the frontier-class open models get the smallest allowance, because OpenCode says it passes on bulk discounts where it has them and has not negotiated them yet on the newest models. Kimi K3 at about 110 requests per five hours is a very different plan from MiniMax M3 at 3,200.
OpenCode's own page says Go is "designed primarily for international users". Only one member per workspace can subscribe, and when you hit a limit you either drop to the free promotional models or let Go draw on a Zen balance. For a US team, that makes Go a good personal plan and an awkward team purchase.
What Claude Code costs
Claude Code has no separate price. It comes with a Claude plan, and the plan sets how much you can use it. Every figure below is from claude.com/pricing and Anthropic's Claude Code cost documentation.
Swipe to see more
| Plan | Price | Usage | What it means |
|---|---|---|---|
| Claude Pro | $17/mo annual, $20 monthly | Baseline allowance | Includes Claude Code in the terminal and IDEs. The quota is shared with Claude chat. |
| Claude Max 5x | $100/mo, monthly only | 5x Pro | No annual price is published, so a year is $1,200. |
| Claude Max 20x | $200/mo, monthly only | 20x Pro | The only individual tier that is cheaper per unit of usage than Pro. |
| Team, standard seat | $20/seat annual, $25 monthly | More than Pro | Sold for teams of 2 to 150, with central billing and SSO. |
| Team, premium seat | $100/seat annual, $125 monthly | 5x a standard seat | Anthropic lists Claude Code and Cowork against this seat type. |
| Anthropic API key | Sonnet 5 $2 in, $10 out per million tokens | Metered, no cap | The same rate OpenCode pays on your key or through Zen. |
The quota is shared between Claude Code and Claude chat, so a morning of research in the browser leaves less for coding. And a stray ANTHROPIC_API_KEY in your shell makes Claude Code bill that key at API rates instead of your plan. The tier-by-tier detail is on our Claude Code Max plan pricing page.
What a month of Claude costs in each tool
Because Claude inside OpenCode is always metered, the comparison is simple arithmetic. Anthropic publishes the number that drives it: Claude Code averages "around $13 per developer per active day" across enterprise deployments, with 90% of users under $30. OpenCode pays the same list rate on your key or through Zen, so the same work costs about the same per token in either harness.
Swipe to see more
| How often you code with Claude | OpenCode, metered | Claude Code, capped plan | What that means |
|---|---|---|---|
| 4 active days a month | About $52 | $20 on Pro, if the cap holds | Pro is cheaper. Tokens win only if Pro keeps stopping you and you would otherwise jump to Max. |
| 8 active days a month | About $104 | $100 on Max 5x | Roughly even. This is the crossover point. |
| 15 active days a month | About $195 | $100 on Max 5x, or $200 on Max 20x | The plan wins by about $95 if Max 5x holds, and breaks even at Max 20x. |
| 21 active days a month | About $273 | $200 on Max 20x | Claude Code on Max 20x saves about $73 a month, close to $900 a year. |
The crossover sits at about eight active days a month. Below it, paying per token through OpenCode costs less than Max 5x. Above it, Claude Code on a Max plan is the cheaper way to buy the same Claude work, and the gap grows every extra day you code. A developer on Claude every working day saves close to $900 a year by using Claude Code on Max 20x instead of Claude through OpenCode.
The dollar figures are our arithmetic on Anthropic's published $13 average and do not include Zen's card fees. Your own average depends on model, context size and how much the agent reads per task, so measure a week with Claude Code's /usage command or your provider dashboard first.
The crossover that does work: Claude Code on OpenCode Go
Your Claude plan cannot go into OpenCode, but Claude Code can go onto OpenCode's plan. OpenCode lists Claude Code as a validated Go client that sends its session header natively, and several Go models, including the MiniMax and Qwen models, are served on an Anthropic-style messages endpoint. Claude Code's own documentation covers pointing it at a gateway with ANTHROPIC_BASE_URL plus a gateway credential.
One detail from Anthropic's gateway page matters here: setting the base URL alone does not replace your subscription. If a claude.ai login is still saved, its limits and billing apply. You need the gateway credential set as well. Done properly, you keep the Claude Code workflow you know, your CLAUDE.md files and your hooks, and swap a $100 or $200 plan for $10 of open-model usage on tasks where an open model is good enough.
Works both ways
Your CLAUDE.md travels
OpenCode reads a project CLAUDE.md when no AGENTS.md exists, plus ~/.claude/CLAUDE.md and ~/.claude/skills, so trying it costs no setup time.
Works one way
Claude Code on the Go meter
Validated by OpenCode for Go. You pay $10 and get open models inside the Claude Code harness.
Does not work
Your Claude plan in OpenCode
Prohibited by Anthropic and removed by OpenCode. Claude in OpenCode means paying per token.
Is OpenCode as good as Claude Code for your situation?
On the same model the two harnesses are close; OpenCode's pull is model freedom and Claude Code's is the tightest Anthropic integration and a plan that caps the bill. So the right choice follows from what you already pay for and how often you code.
You code with Claude most working days
Claude Code on Max
Claude inside OpenCode is billed per token. At Anthropic's $13 per active day average, a daily user runs about $273 a month on tokens, against a $200 cap on Max 20x.
You already pay for ChatGPT Plus or Pro
OpenCode on your ChatGPT plan
OpenCode signs in to ChatGPT directly, so the harness costs nothing extra. Codex does the same job on the same plan, which is worth comparing before you switch.
Your company pays for GitHub Copilot
OpenCode on Copilot
OpenCode accepts a Copilot login, so the seat your company already bought drives a terminal agent with no new invoice. Some models need a Pro+ plan.
You want the lowest monthly bill
OpenCode Go at $10
Open models at a fixed $10 with 5-hour, weekly and monthly limits, plus the option to fall back to a Zen balance. Test quality on real tickets before you commit a team.
You like Claude Code but want cheaper models
Claude Code on OpenCode Go
OpenCode lists Claude Code as a validated Go client, and several Go models are served on an Anthropic-style endpoint. You keep the harness and swap the meter.
You want finished pull requests, not a pair session
A repo agent on a flat bill
Both tools sell time at the keyboard. If the job is a described task returned as a tested pull request on GitHub or GitLab, with no key and no meter to manage, that is a different purchase.
OpenCode Enterprise vs Claude Team and Enterprise
For a team the question becomes governance plus a meter, and the two vendors package them in opposite ways. OpenCode Enterprise is priced per seat on a custom quote and covers the control layer: a central config for the whole organization, SSO integration, and a lock that sends every request through your internal AI gateway. If you bring your own gateway, OpenCode says it does not charge for tokens. Zen team workspaces, with admin and member roles and per-member monthly limits, are free while in beta.
Anthropic sells the other shape. Claude Team runs from 2 to 150 people with capped seats at $25 or $125 a month, which keeps the bill predictable. Claude Enterprise charges a seat fee for access and bills all usage at API rates. So a team on OpenCode Enterprise with Claude behind its gateway, and a team on Claude Enterprise, both end up paying tokens; only Claude Team keeps a hard cap. We priced that decision for an eight-person team in what Claude Code for teams really costs.
Where we fit, and where both tools beat us
The losses first. OpenCode and Claude Code are both strong interactive agents. They sit in your terminal, hold a conversation about your code, and suit exploratory work where you do not yet know exactly what you want built. OpenCode's model freedom is something we do not offer, and Claude Code's reach from terminal to IDE to CI is broader than ours.
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 and no invoice that moves with how busy the sprint was: it is $29 a month billed yearly.
If you are comparing meters because the bill is the problem, 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 Claude Code: the questions people search
Is OpenCode as good as Claude Code?
On the same Claude model, the gap is in the harness, not the model: Claude Code has deeper Anthropic integration, while OpenCode runs 75+ providers and local models. The practical difference is billing. Claude Code can use your Claude plan and OpenCode cannot, so a heavy Claude user pays more for the same work in OpenCode.
Can I use OpenCode with Claude subscription?
Not in current versions. OpenCode's own provider docs say plugins that use Claude Pro or Max plans exist, that Anthropic explicitly prohibits this, and that OpenCode stopped bundling them in version 1.3.0. Anthropic's terms say sign-in with a Claude plan is for Claude Code and its own apps. To use Claude in OpenCode, use an API key or Zen.
Can I use OpenCode Go in Claude Code?
Yes. OpenCode lists Claude Code as a validated Go client that sends its session header natively, and several Go models, including MiniMax and Qwen models, are served on an Anthropic-style messages endpoint. Claude Code reaches a gateway through the ANTHROPIC_BASE_URL setting plus a gateway credential. You keep Claude Code and pay OpenCode's $10 instead of a Claude plan.
Can I use OpenCode with ChatGPT subscription?
Yes. OpenCode's OpenAI provider has a ChatGPT Plus/Pro sign-in option that opens your browser to authenticate, and its docs recommend a ChatGPT plan for OpenAI models. It also accepts GitHub Copilot and GitLab Duo subscriptions with no extra setup. Claude plans are the major subscription it does not accept.
How much is OpenCode Go?
OpenCode Go costs $10 a month and can be cancelled any time. It covers 27 models, mostly open ones plus GPT 5.6 Luna and Grok 4.6, and gives each a monthly dollar allowance of $15, $30 or $60. You can spend 20% of that in any 5-hour window and 50% in a week.
How much does OpenCode Zen cost?
Zen is pay as you go per million tokens at the model's list rate: Claude Sonnet 5 is $2 in and $10 out, Opus 5 is $5 and $25. OpenCode passes card fees through at 4.4% plus $0.30 per top-up, and by default reloads $20 when your balance drops below $5. You can set a monthly limit.
OpenCode Zen vs Claude Code pricing: which is cheaper?
For Claude models, Zen charges the same list rate as Anthropic's API plus card fees, so it is never cheaper than a Claude plan once you code most days. At Anthropic's average of $13 per active day, Zen passes Max 5x at about eight days a month. Zen is cheaper only for light use or for non-Claude models.
Is OpenCode Go worth it?
It is worth it if open models are good enough for your work, because $10 buys up to $60 a month of usage on many of them. It is not a Claude substitute and holds no Claude models. Its own page says it is designed primarily for international users, so US teams should test latency and quality on real tickets first.
Does OpenCode have limits?
The software itself has none; limits come from how you pay. Go caps each model at a monthly dollar amount, with 20% available per 5 hours and 50% per week. Zen lets an admin set a monthly limit for the workspace and each member. On your own API key, only your provider's limits apply.
Does OpenCode read CLAUDE.md?
Yes, as a fallback. OpenCode reads AGENTS.md first, and uses a project CLAUDE.md if no AGENTS.md exists, plus ~/.claude/CLAUDE.md and ~/.claude/skills globally. That makes moving from Claude Code cheap in setup time. You can switch the behavior off with the OPENCODE_DISABLE_CLAUDE_CODE environment variable.
What does OpenCode Enterprise cost?
OpenCode prices Enterprise per seat on a custom quote and does not publish a figure. If you route developers through your own LLM gateway, it does not charge for tokens. The package is a central config, SSO integration and a lock that sends every request to your approved gateway.
Is OpenCode completely free?
The software is free and MIT-licensed, but every task calls a model, and that is where the bill is. You pay your provider directly, top up Zen, subscribe to Go at $10, or reuse a ChatGPT or Copilot plan. Only a handful of limited-time promotional models on Zen cost nothing, and some of those may use your data.
Keep researching
The Claude side in more depth: Claude Code pricing across every plan, how the Claude Code usage limits work and the Claude Code CLI.
Other open-source harnesses against the vendor agents: Cline against Claude Code, Cline against Codex, OpenCode against Codex on a ChatGPT plan, Codex against Claude Code and Cursor against Claude Code.
A repo agent with no meter to pick
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 usage cap, and nothing merges without you.