dalao.ai

GitHub Copilot

GitHub's AI coding agent, from the open issue to the merged pull requestReference profile

About GitHub Copilot

GitHub Copilot is GitHub's AI coding assistant and agent. It covers editor completions, chat, agent mode, cloud tasks and code review, and can take a change from an open issue to a mergeable pull request. Since 2026-06-01 every plan bills by usage in GitHub AI Credits, and at Universe on 2026-10-28 GitHub made Copilot Fusion generally available, which picks and combines models per task. As of 2026-10-31 there are four individual plans, from free to $100 a month.

EditorialUpdated 5 min read

What problem it solves

Writing code is only one part of shipping software. Learning an unfamiliar codebase, breaking down a task, editing many files, opening a PR and reviewing it all take time. Copilot puts these steps inside the GitHub workflow developers already use. In the docs' words, a change can start as an issue, be picked up by an agent, come back as a pull request, get reviewed for quality and security, and be merged.

The docs group its features into four kinds: assistive (completions and answers while you work), agentic (you give a goal and it researches the repo, plans, edits files and runs tools), customization (instructions, prompt files, skills and custom agents), and outside agents, models and MCP tools working alongside it.

GitHub Copilot is an AI coding agent that helps you write, understand, and ship software.

GitHub Copilot pricing page FAQ (GitHub's own description)

Core capabilities

  • Completions and chat: inline suggestions and next edit suggestions in the IDE are unlimited on paid plans and don't use AI Credits. Free gets 2,000 completions a month.
  • Agent mode: runs in VS Code, Visual Studio, JetBrains, Eclipse and Xcode, and connects to MCP servers.
  • Copilot in the cloud: assign an issue and Copilot researches, plans, writes code and opens a PR in the cloud. Pro+ and Max can also hand tasks to third-party coding agents such as Anthropic's Claude and OpenAI Codex (preview).
  • Code review: PR reviews on GitHub. Universe added a Max level, with Lite and Balanced for everyday reviews.
  • Copilot CLI and the Copilot app, for the terminal and the desktop, on every plan.
  • Models from several vendors: the pricing page lists models from Anthropic, OpenAI, Google, xAI, Microsoft and Kimi.
  • Copilot Fusion, generally available since 2026-10-28. You pick an effort level (Standard, High or Max) and the system decides which models to use. It either runs one model, lets a cheaper model draft and escalates after a quality check (Cascade), or has one model draft and another review (Critique). The same day, routing between local and cloud models shipped as an experimental preview.

Who uses it

GitHub's own pages give no count of Copilot users or paying customers. The Fusion changelog quotes Hari Pokhrel, a principal software engineer at AT&T, who says Fusion Standard improved token efficiency across AT&T by picking the model for each task. That is a customer quote GitHub chose to publish.

The pricing page lists VS Code, Visual Studio, JetBrains IDEs, Xcode, Neovim, Eclipse, Zed, Raycast and SQL Server Management Studio as places to use it, along with github.com and GitHub Mobile.

How it compares

Like Claude Code, Copilot runs in the terminal, the IDE, a desktop app and the browser (Claude Code's docs: "Available in your terminal, IDE, desktop app, and browser"). The difference is that Copilot is built into the GitHub platform, so issues, PRs, code review, Actions and org policies are one system. Its models also come from several vendors, and admins can switch each one on or off.

In the Fusion changelog, GitHub says Fusion Max scored 60 on the Artificial Analysis Coding Agent Index, within two points of Claude Code with Opus 5.5 Max, at about half the average API cost per task and about 40% less time. That comparison is GitHub's own; see the post for the test setup.

Who it's for / Not for

Good fit:

  • Individuals and teams whose code and collaboration already live on GitHub.
  • Companies that need central control over models, budgets, content exclusion and audit logs.
  • Developers who want several vendors' models in one subscription without switching by hand.

Not a fit:

  • Teams whose code lives only on GitHub Enterprise Server: Copilot isn't integrated into GitHub Enterprise Server, and licenses can only be assigned on GitHub.com or GHE.com. Developers can still use it in local clients such as Copilot CLI through a cloud enterprise account, and an air-gapped Copilot CLI setup is in technical preview.
  • Users who need a fixed monthly bill and don't want to manage usage budgets.
  • Teams that require open-source or self-hosted tools.

Getting started

  1. As an individual, turn on Copilot Free at github.com/settings/copilot or upgrade to a paid plan. Students, teachers and maintainers of popular open source projects may get paid features for free.
  2. Install Copilot in your IDE and sign in with your GitHub account. Organization members need an admin to assign them a seat.
  3. To try Fusion, choose Copilot Fusion in the model picker and pick an effort level. It works in VS Code, the Copilot app, Copilot CLI and JetBrains IDEs.

Pricing & license

As of 2026-10-31, one GitHub AI Credit is $0.01, drawn down at each model's token prices:

  • Free: $0, 2,000 completions a month and limited chat and agent use. Models are auto-selected only; the pricing page ticks just two entry-level models, Claude Haiku 4.5 and GPT-5 mini.
  • Pro: $10 a month, with 1,000 base credits plus a 500-credit flex allotment, 1,500 in total.
  • Pro+: $39 a month, 7,000 in total (3,900 base plus 3,100 flex).
  • Max: $100 a month, 20,000 in total (10,000 each).
  • Business: $19 per seat a month with 1,900 credits per user, pooled across the org. Enterprise: $39 per seat with 3,900 per user.
  • Copilot access: $0 per seat for GitHub Enterprise Cloud license holders, or $10 per user a month standalone, with no included credits and metering from the first use. For now it is set up through a GitHub or Microsoft account team; self-serve purchase isn't open yet.

Credits reset at midnight UTC on the first of each month and don't roll over, and GitHub says the flex allotment may change over time. Past the allowance you can set a budget and keep paying $0.01 per credit. Code review also uses GitHub Actions minutes. Fusion bills each model it uses at that model's standard rate, with a limited-time 30% discount on paid plans.

License: Copilot is a proprietary commercial service. The source of the Copilot Chat extension for VS Code was published under MIT in microsoft/vscode-copilot-chat; that repo is now archived and development has moved into microsoft/vscode.

Risks & limits

  • Costs are harder to predict. Since 2026-06-01 billing is by tokens instead of counted premium requests, and once credits run out there is no fallback to a cheaper model. Extra usage on individual plans may be capped.
  • Data use: the pricing FAQ says prompts, outputs and code snippets from Free, Pro and Pro+ users may be used to train and improve models unless they opt out in settings. Business and Enterprise data is handled under the data protection agreement.
  • The benchmark figures come from GitHub's own changelog.

Our verdict

For teams whose code already lives on GitHub, Copilot's strength is how deep the integration goes: issues, agents, PRs and review sit on one platform, and admins manage models and budgets in one place. With token billing since 2026, heavy use costs depend on which models you pick and how long agent runs take, so it's worth watching the usage dashboard on Free or Pro before choosing a plan. Fusion removes the chore of picking models by hand; so far the only performance numbers are GitHub's.

Why it matters

GitHub Copilot was one of the first widely adopted AI coding assistants and is built into major editors and GitHub itself.

Sources

Media & third party

  1. GitHub Docs: GitHub Copilot plansGitHub Docs · Accessed Oct 31, 2026
  2. GitHub Docs: What is GitHub Copilot?GitHub Docs · Accessed Oct 31, 2026
  3. GitHub Docs: Usage-based billing for individualsGitHub Docs · Accessed Oct 31, 2026
  4. GitHub Blog: GitHub Copilot is moving to usage-based billinggithub.blog · Accessed Oct 31, 2026
  5. GitHub Blog: GitHub Universe 2026: Evolving every layer of the stackgithub.blog · Accessed Oct 31, 2026
  6. GitHub Docs: About GitHub Copilot on GitHub Enterprise ServerGitHub Docs · Accessed Oct 31, 2026

Related products

  1. GitHub Copilot plans and pricingGitHub · Accessed Oct 31, 2026
  2. GitHub Changelog: Updates to GitHub Copilot billing and plansgithub.blog · Accessed Oct 31, 2026

Updates & reading

NewsGitHub Changelog

GitHub makes Copilot Fusion generally available at Universe 2026, routing and orchestrating models per task

At Universe 2026 on Oct 28, GitHub made Copilot Fusion generally available: users choose Fusion and a Standard, High or Max effort level, and Copilot picks and combines models for each task. GitHub says Fusion Max scored 60 on the Artificial Analysis Coding Agent Index, within two points of Claude Code with Opus 5.5 Max, at about half the cost per task.

2 sources