# Live cost: what every turn costs, as it runs > See what each Claude Code or Codex turn costs while it runs, in every chat, plus totals by source and project on the Analytics page. Source: https://shuriken.sh/features/live-cost See what each Claude Code or Codex turn costs while it runs, in every chat, plus totals by source and project on the Analytics page. By Shuriken Updated 27 October 2026 5 min read **Contents** Shuriken is a Claude Code cost tracker per session that runs inside every chat: the turn under way shows its tokens and cost while it runs, for Claude Code, Codex and the other agents that report usage as they go. The figure sits beside the turn's running time, next to the fork button, and climbs as the agent works. When the turn ends it adds to the chat's total. Analytics, a page under Home, adds it all up by source, project and model, across every computer you've linked. ## Why we built it # Our own setup is the one we kept hearing about: more than one agent on more than one plan. Claude Code on a subscription, Codex on a ChatGPT plan, an API key for when both are out. Each counts usage its own way, and none of them gives one running figure across all of them. After a long turn the question was always whether that was a small job or a big one, and the answer came later, in a console, after it mattered. We put the meter beside the chat instead, where it's visible while the turn can still be steered. ## What you see # While a turn runs, its footer shows how long it has been going, then how many commands it ran, the lines it added and removed, and its tokens and cost so far, like "42k / $0.35". The footer sits next to the fork button and the numbers grow as the agent reads, thinks and edits. Subagents and workflows the turn starts count too. Nothing of it sits in the message box. When it ends, the turn joins the chat's total, which the context-window readout shows as "Chat cost" and the amount. Hover a money figure and its note says what it is. On a subscription that's "API-equivalent cost: what this usage would cost at API prices. You pay for your subscription, not this." On a key it's "Spent on your API key." No setting to turn on, and no pop-up. ## More than Claude # Each agent reports usage its own way, and getting a trustworthy number out of each one was the hard part. - Claude Code streams each request's token counts. Shuriken adds them up and prices them at Anthropic's list rates. - Codex reports a running token total, and Shuriken prices the difference since the turn began. A rewind or a fork starts fresh, so it never measures against an old total. - Gemini CLI writes each call's tokens to its session file. Shuriken reads it when the turn ends and prices each call by its own model, so a Gemini turn's cost shows once it finishes. - Pi and other agents that report usage during a turn are read the same way. An agent that reports only at the end of a turn shows its cost then. When we can't price something, we don't guess. A turn where one call used a model with no published price shows its tokens and no price. An agent that reports zero because it doesn't know reads as unknown, not as free. [Local models](https://shuriken.sh/features/local-models) are the exception: a turn at a local address, such as Ollama or LM Studio, isn't priced and reads "42k / local". We don't call it free, because a local address can also forward to a paid service. ## What the Claude Code cost figure means # It is an API-equivalent cost: the turn's tokens priced at the provider's published list rates. On an API key, that's close to what you'll pay, and the provider's bill is still the authority. Anthropic says the same about Claude Code's own session figure: it is computed locally at list price, and it is an estimate ([Claude Code docs](https://code.claude.com/docs/en/costs), read 22 October 2026). On a subscription, you don't pay that amount. Anthropic notes that for Pro and Max subscribers the session cost isn't relevant for billing. It is still the quickest sense of how heavy a turn was, which matters when a 5-hour window is half gone. To compare the two with your own numbers, [API key or subscription?](https://shuriken.sh/guides/claude-code-api-vs-subscription) walks through the sums. ## Totals by account and project # The per-turn line is for right now. Analytics, under Home, answers "where did this week go?" with: - spend for today, the last 7 and 30 days, and all time; - daily spend, split by model, agent or project; - spend by source, each subscription and API key on its own; - spend by project and by model. Each turn counts against the source it actually ran on. If [source switching](https://shuriken.sh/features/source-switching) moved a chat from one account to another halfway, each account carries its own part. Money shows in US dollars or your own currency, converted at the day's rate. The screens are in [Spend by account and project, on one screen](https://shuriken.sh/features/fuel). ## What it's good for # A "quick fix" still climbing after a few minutes is a runaway turn. The meter shows it early, while it's still cheap to steer or stop, and the Steer button sits on each row of the [queue](https://shuriken.sh/features/queue). The meter also makes API keys less scary. Switch to API keys (Sources › Switch settings) is off until you turn it on. Then API keys take chats only once every subscription of that agent is out, and Spend cap per fallback, $25 unless you change it, stops the run there, counted across every key it uses: "Past this, chats wait for a subscription". Watching a few turns tells you what a cap really buys. To pick between models with your own numbers, run the same kind of task on two models and compare, from your own work. We have no savings figure to quote you. Seeing the meter doesn't make a turn cheaper by itself. It puts the information next to the decision. ## Other tools that count cost # Fair note: you don't need Shuriken to see what Claude Code costs. Claude Code itself shows a session's estimated cost in /usage, can put it in its status line, and shows a live count of the current turn's output tokens beside its spinner ([Claude Code docs](https://code.claude.com/docs/en/costs)). ccusage is an MIT-licensed command-line tool that reads the usage logs Claude Code, Codex, OpenCode, Gemini CLI and other agents write on your computer, and turns them into daily, monthly and per-session reports of tokens and estimated cost, without uploading your data. It also tracks Claude's 5-hour blocks and has a status line for Claude Code, in beta ([ccusage](https://ccusage.com/), [GitHub](https://github.com/ccusage/ccusage), both read 22 October 2026). In one terminal on one computer, those may be all you need. Shuriken puts the number somewhere else: live in every chat for the agents that report usage as they go, with totals across every computer you've linked, each turn counted against the source it ran on. Live cost is on every plan, Free included. To see it on your own agents, [download Shuriken](https://shuriken.sh/download). Share [X](https://x.com/intent/post?url=https%3A%2F%2Fshuriken.sh%2Ffeatures%2Flive-cost&text=Live%20cost%3A%20what%20every%20turn%20costs%2C%20as%20it%20runs)[LinkedIn](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fshuriken.sh%2Ffeatures%2Flive-cost)[Hacker News](https://news.ycombinator.com/submitlink?u=https%3A%2F%2Fshuriken.sh%2Ffeatures%2Flive-cost&t=Live%20cost%3A%20what%20every%20turn%20costs%2C%20as%20it%20runs)[Reddit](https://www.reddit.com/submit?url=https%3A%2F%2Fshuriken.sh%2Ffeatures%2Flive-cost&title=Live%20cost%3A%20what%20every%20turn%20costs%2C%20as%20it%20runs)[E-mail](mailto:?subject=Live%20cost%3A%20what%20every%20turn%20costs%2C%20as%20it%20runs&body=https%3A%2F%2Fshuriken.sh%2Ffeatures%2Flive-cost) ## Sources - [Claude Code docs: Manage costs effectively](https://code.claude.com/docs/en/costs) (accessed 22 October 2026) - [ccusage](https://ccusage.com/) (accessed 22 October 2026) - [ccusage on GitHub](https://github.com/ccusage/ccusage) (accessed 22 October 2026) ## Frequently asked questions **Is the number what I pay on a Claude subscription?** No. On Pro or Max your usage is part of the plan. The figure is what the same tokens would cost at API list prices, a fair measure of how heavy a turn was. Shuriken never bills model usage. **How accurate is it on an API key?** Close, because it prices the tokens the agent reports at the provider's published rates. Your provider's bill is still the authority, as it is for Claude Code's own estimate. **Which agents does it work with?** Claude Code, Codex and other agents that report usage as they go show it while the turn runs. Gemini CLI reports its tokens only when a turn ends, so its cost shows then. When part of a turn has no published price, Shuriken shows no price rather than a guess. **What about local models?** Turns on a model at a local address, such as Ollama or LM Studio, are not priced and read local. We don't call them free, because a local address can also forward to a paid service. **Is live cost on the Free plan?** Yes. Live cost and the Analytics page are part of every plan. ## Try Shuriken on your own computer Add your Claude, Codex and Gemini accounts once. Shuriken switches between them so your agents keep working. [Download](https://shuriken.sh/download) [Pricing](https://shuriken.sh/pricing) Coming soon ## Keep reading - [Best AI coding subscription in 2026, by how you work: Comparison 27 Oct 2026 5 min read The best AI coding subscription in 2026 depends on how you work. Claude, ChatGPT, Google, Cursor and Copilot plans compared: price, limits, who each suits.](https://shuriken.sh/compare/best-ai-coding-subscription) - [Claude Code alternatives: the terminal agents worth trying: Comparison 27 Oct 2026 10 min read Which Claude Code alternative fits depends on your plan: Codex on ChatGPT, Copilot CLI on any Copilot plan, OpenCode for any model. 13 agents, sourced.](https://shuriken.sh/compare/claude-code-alternatives) - [ChatGPT Plus vs Claude Pro for coding: the $20 plans: Comparison 27 Oct 2026 7 min read ChatGPT Plus and Claude Pro both cost $20 and include a coding agent, Codex or Claude Code. Their limits, models and extras, from each vendor's own pages.](https://shuriken.sh/compare/chatgpt-plus-vs-claude-pro-for-coding) - [Claude Max vs ChatGPT Pro for coding (2026): Comparison 27 Oct 2026 6 min read Claude Max vs ChatGPT Pro for coding: both start at $100 a month. Max has a five-hour window plus a weekly limit; Pro has no five-hour limit for now.](https://shuriken.sh/compare/claude-max-vs-chatgpt-pro-for-coding)