caprock

GitHub Copilot vs Claude Code: pricing, AI credits and what heavy use costs

Pick GitHub Copilot for completions as you type and several makers' models from $10 a month; pick Claude Code if you run an agent most of the day, because a Claude plan is a flat fee while Copilot bills agent work per token past its included AI credits. Claude Code starts at $20 a month. Read 10 October 2026.

Dmitriy Solodukha · checked against Caprock v0.95.0

Copilot Pro, a month
$10
Copilot Pro, a month
includes $15 of AI credits
Claude Pro, a month
$20
Claude Pro, a month
$17 billed annually · the entry plan for Claude Code
a median day of my Claude work at API rates
$153
a median day of my Claude work at API rates
Estimate · 56 days, 18 Jul – 2 Oct 2026 · Caprock does not read Copilot
of credits in Copilot Max, a month
$200
of credits in Copilot Max, a month
$100 plan

Caprock reads Claude Code, not Copilot. Every Copilot figure here is published by GitHub; the $153 is my Claude Code work at Anthropic's list prices, which Copilot's Claude rates match — an estimate, not a Copilot bill.

The short answer

Pick GitHub Copilot or Claude Code

Pick GitHub Copilot if…

  • You want code completions in the editor: they are unlimited on every paid Copilot plan and do not use credits.
  • You want one subscription for several makers' models — Claude, GPT and Gemini — in VS Code, on github.com and in a CLI.
  • Your team lives on GitHub: Copilot's cloud agent works from issues and opens pull requests.

Pick Claude Code if…

  • You run an agent most of the day: a Claude plan is a flat fee inside its limits, while Copilot bills agent work per token past the included credits.
  • Heavy use: my median Claude Code day came to $153 at list prices — two such days pass a month of Copilot Max's $200 of credits.
  • You want Claude through Amazon Bedrock, Google Cloud or Microsoft Foundry, or Claude Code's hooks, subagents and status line.

Facts about price, limits and where each runs, read 10 October 2026. Not a verdict on which writes better code.

Side by side

GitHub Copilot vs Claude Code, side by side

GitHub Copilot and Claude Code, from each one's documentation, read 10 October 2026
GitHub CopilotClaude Code
Made byGitHub (Microsoft)Anthropic
What it isAn assistant in your editor and on GitHubAn agent for the terminal and IDEs
Where it runsIDEs, github.com, CLI, cloud agentTerminal, IDE extensions, desktop app, web
ModelsAnthropic, OpenAI, Google and othersClaude only
Open sourceNoNo
Cheapest way inFree; Pro, $10 a monthClaude Pro, $20 a month
BilledAI credits per token, past includedPlan fee in usage limits, or per token
InstallIDE extension; CLI by npm, Homebrew, WinGetInstaller, npm, Homebrew, WinGet, apt, dnf, apk
Read by CaprockNoYes, every session
Completions as you typeYes, unlimited on paid plansNo
MCP, custom agents, hooksYes, in the CLI and cloud agentYes

Plans, per month

GitHub Copilot pricing vs Claude Code pricing

Monthly price of each plan on one axis. Business, Enterprise and Team are per user.

GitHub Copilot

  • Copilot Free$0limited chat and agent use
  • Copilot Pro$10$15 of credits
  • Copilot Business$19a user · $19 of credits
  • Copilot Pro+$39$70 of credits
  • Copilot Enterprise$39a user · $39 of credits
  • Copilot Max$100$200 of credits

Claude Code

  • Claude Pro$20$17 billed annually
  • Team, standard seat$25$20 billed annually
  • Claude Max 5×$1005× Pro's usage
  • Team, premium seat$125$100 billed annually
  • Claude Max 20×$20020× Pro's usage
  • Claude Free — no Claude Code
  • Enterprise — $20 a seat + usage at API rates
  • API key — per token

Both columns share one axis, $0 to $200 a month. Read 10 October 2026 from github.com/features/copilot/plans, GitHub's billing docs and claude.com/pricing. One AI credit is $0.01. The included credits count a flex allotment ($5, $31 and $100 on Pro, Pro+ and Max) that GitHub says may change.

Usage limits

Usage limits: monthly credits against a five-hour window

How GitHub Copilot and Claude Code cap use, read 10 October 2026
GitHub CopilotClaude Code
How usage is cappedAI credits a monthFive-hour window and a weekly limit
Size publishedYes, in dollarsOnly relative: 5× or 20× Pro
Weekly limitNone publishedYes, across all models
Past the limitA budget, at $0.01 a creditWait, usage credits, or an API key
Resets1st of the month, 00:00 UTCEvery five hours; weekly
Not countedCompletions, next edit suggestions—
Copilot charges input, output and cached tokens, converted to credits. GitHub says additional usage on individual plans “may be capped”.

Per million tokens

What Claude costs inside Copilot

Per million tokens, from GitHub's models and pricing table. Its Claude rates match Anthropic's list.

Claude in Copilot prices per million tokens, read 10 October 2026
ModelInputCache writeCache readOutput
Claude Opus 5.5$4$5$0.20$20
Claude Opus 5$5$6.25$0.50$25
Claude Sonnet 5.5$2$2.50$0.10$10
Claude Haiku 5.5$0.10$0.125$0.01$0.50
Claude Haiku 4.5$1$1.25$0.10$5
Claude Fable 5.1$10$12.50$0.25$50
Read 10 October 2026 from docs.github.com. Opus 4.8 costs the same as Opus 5; those two ran most of my measured calls. Haiku 5.5 has a long-context row at 5× these rates.

Measured · read 3 October 2026

The Claude Code side, measured

Caprock reads Claude Code's transcripts, not Copilot. These are my Claude Code calls at list prices; the Copilot column stays empty.

GitHub Copilot
Not measured — Caprock does not read Copilot.
Claude Code
79,014 calls on 56 active days, 18 Jul – 2 Oct 2026mostly Opus 5 and Opus 4.8

Median cost per call

  • GitHub CopilotCaprock does not read Copilotnot measured
  • Claude Code18 Jul – 2 Oct 2026$0.094
A median Opus 5 call alone: $0.112.

Median active day

  • GitHub CopilotCaprock does not read Copilotnot measured
  • Claude Code18 Jul – 2 Oct 2026$153
56 active days came to $13,922 at list prices.

Median context per call

  • GitHub CopilotCaprock does not read Copilotnot measured
  • Claude Code18 Jul – 2 Oct 2026169K
Almost all of it read from the cache. Copilot's agent builds its own context, so its calls will not match these.

Where the money went

  • GitHub CopilotCaprock does not read Copilot
  • Claude Code18 Jul – 2 Oct 2026
ShareClaude Code
Cache read79%
Cache write16%
Output5%
Fresh inputunder 0.1%
Share of the list-price cost. Most of my cache writes were the one-hour kind, which costs more than the one write rate GitHub lists.

Same Claude work, two bills

One heavy day at API rates, against a month of Copilot credits

EstimateCaprock does not read Copilot

  • A median day, Claude at API rates$153estimate, per day
  • Copilot Pro, credits$15a whole month
  • Copilot Pro+, credits$70a whole month
  • Copilot Max, credits$200a whole month
Estimate — Caprock does not read Copilot. The day is my median active day of Claude Code calls at Anthropic's list prices, mostly Opus 5 and 4.8, which GitHub lists at the same rates. Copilot's agent builds its own context, so its calls would not match these one for one.

More comparisons

The other agents, against Claude Code

Your own machine

See your Claude side.

Caprock is a free, open-source desktop app that keeps the ledger of every agent it reads, Claude Code, Codex, OpenCode and Gemini CLI, the sessions you start in your own terminal included: who is waiting on you, who is stuck, how close each plan is to its limit, and each pull request with what it cost, at list price. It does not read Copilot. Local; no account.

Download for Mac

Notarized by Apple. One file for Apple silicon and Intel, macOS 13 or later.

Or run it yourself:

curl -fsSL https://caprock.dev/mac.sh | sh

Paste in Terminal: it installs Caprock and opens it.Or with Homebrew: brew install --cask dspv/tap/caprock-app

The full guide

GitHub Copilot vs Claude Code, step by step

GitHub Copilot is GitHub's assistant: completions as you type, chat and an agent in your editor, an agent on github.com and a CLI, on models from Anthropic, OpenAI, Google and others. Claude Code is Anthropic's agent and runs Claude only. Since 1 June 2026 Copilot bills chat and agent work in AI credits, charged per token; a Claude plan charges a flat fee with usage limits. Claude in Copilot costs the same per token as Anthropic's list.

GitHub Copilot pricing vs Claude Code pricing

Copilot's individual plans are Free, Pro at $10 a month, Pro+ at $39 and Max at $100; Business is $19 a user and Enterprise $39. Each paid plan includes AI credits worth a set amount — $15 on Pro, $70 on Pro+, $200 on Max — at one cent a credit, reset on the 1st of the month. Code completions and next edit suggestions do not use credits and stay unlimited on paid plans.

Past the included credits you set a budget and pay a cent a credit, or wait for the month to reset; GitHub says extra usage on individual plans "may be capped". Part of each plan's credits is a flex allotment that GitHub says may change. Before June, plans counted premium requests instead; annual subscribers keep that system until their plan ends.

Claude Code is included in every paid Claude plan, from Pro at $20, with a five-hour window and a weekly limit instead of a monthly pool. The details: Claude Code pricing.

What Claude costs inside Copilot

GitHub's models and pricing table lists Claude Opus 5.5 at $4 per million input tokens and $20 output, Sonnet 5.5 at $2 and $10, Haiku 5.5 at $0.10 and $0.50, with cache reads and writes charged too — the same as Anthropic's list. So Claude in Copilot's agent costs what it would on an API key, paid from your credits first.

What heavy agent work comes to can only be estimated here. Caprock does not read Copilot; the estimate is my own Claude Code calls at Anthropic's list prices, mostly Opus 5 and 4.8, which GitHub lists at the same $5 and $25. A median active day came to $153, over 56 active days from 18 July to 2 October 2026. On Copilot that is about ten times Pro's monthly credits in one day. Copilot's agent builds its own context, so its calls would not match mine one for one; this is not a measured Copilot bill.

Where each one runs

Copilot runs in your editor, on github.com, in GitHub Mobile and as Copilot CLI, which every plan includes:

  1. 1

    Install Copilot CLI with npm (Node.js 22 or later)

  2. 2

    Or with Homebrew

Its cloud agent takes an issue, works in a GitHub Actions environment and opens a pull request, billed in Actions minutes and credits. Copilot CLI and the cloud agent support MCP servers, custom agents, hooks and skills.

Claude Code runs in a terminal, in VS Code and JetBrains, in a desktop app and at claude.ai/code. Its cloud sessions also clone a GitHub repository and open pull requests.

Claude Code inside GitHub

On github.com, "Claude by Anthropic" is one of two third-party coding agents, with OpenAI Codex, in public preview: assign it an issue or mention it in a pull request, and it is billed in Actions minutes and Copilot credits. GitHub's plans page shows third-party agents on Pro+ and Max; its docs also list Pro, Business and Enterprise. Either way it runs on your Copilot plan, not your Claude plan.

Which one to pick

  • Copilot — you want completions as you type and a mix of models in one subscription, with moderate agent use. At $10 it is the cheaper way in.
  • Claude Code — you run an agent for hours a day. Past Copilot's included credits, Claude is billed per token at API rates; a Claude plan is a flat fee inside its limits.
  • Both — many people keep Copilot for completions and run Claude Code for agent work; each goes to its own bill.

Cursor vs Claude Code and Codex vs Claude Code compare the same question for the other agents.

See your Claude side

Caprock reads Claude Code's transcripts from disk, whether or not it started the session, and prices every call — per session, per project, per model — a list-price estimate as Claude Code's own /usage is, next to the plan's dollars and the share of the weekly limit the work used. It does not read Copilot; that stays on GitHub's usage page.

Free, open source; nothing leaves your machine. Install Caprock and it reads back every Claude Code session still on disk.

Questions people ask

Is GitHub Copilot cheaper than Claude Code?
To start, yes: Copilot Pro is $10 a month, Claude Pro $20. For heavy agent use Copilot bills per token past its included credits, while a Claude plan is a flat fee inside its limits.
How much is GitHub Copilot?
Free; Pro $10 a month with $15 of AI credits; Pro+ $39 with $70; Max $100 with $200; Business $19 and Enterprise $39 a user. One credit is $0.01.
Can I use Claude in GitHub Copilot?
Yes. Copilot lists Claude models such as Opus 5.5, Sonnet 5.5 and Haiku 5.5, billed in AI credits at the same per-token rates as Anthropic's list.
What are GitHub Copilot AI credits?
Since 1 June 2026 Copilot bills chat and agent use in AI credits worth $0.01 each, charged per input, output and cached token at each model's rate. Code completions do not use them.
Can Caprock track GitHub Copilot usage?
No. Caprock reads Claude Code's transcripts, not Copilot; it shows what your Claude work costs at API rates against your plan.
Which is better, GitHub Copilot or Claude Code?
Neither across the board. Copilot suits you if you want completions as you type and several makers' models from $10 a month; Claude Code if you run an agent most of the day on a flat Claude plan. This page compares price and limits, not code quality.

Sources

Try it

The record of every coding agent you run, the sessions it did not start included. Free, open source; nothing leaves your machine.

Download for Mac

Notarized by Apple. One file for Apple silicon and Intel, macOS 13 or later.

Or run it yourself:

curl -fsSL https://caprock.dev/mac.sh | sh

Paste in Terminal: it installs Caprock and opens it.Or with Homebrew: brew install --cask dspv/tap/caprock-app