GitHub Copilot

Learning Objectives

• Understand what GitHub Copilot offers across its product surface: completions, chat, edits, agent mode, the async coding agent, and the new agent-native desktop app
• Compare Copilot's pricing tiers and determine which plan fits your workflow
• Evaluate when to use GitHub Copilot versus alternatives like Claude Code, Cursor, or Gemini CLI

What Is GitHub Copilot?

GitHub Copilot is GitHub's AI-powered coding assistant. Originally launched in 2021 as an autocomplete tool, it has evolved into a full suite of AI coding capabilities — inline code completions, conversational chat, multi-file editing, autonomous agent mode, and a background coding agent that can work on GitHub Issues independently.

Copilot is deeply integrated into the GitHub platform. It can review pull requests, work on assigned issues, run in GitHub Actions, and connect to external services through MCP (Model Context Protocol) servers. With over 1.8 million paid subscribers and 77,000+ organizations, it is the most widely adopted AI coding tool in the industry.

What sets Copilot apart from standalone AI coding tools is its multi-model architecture — users can choose from models by OpenAI, Anthropic, Google, and xAI, or let Copilot automatically select the best model for each task.

Pricing

As of June 1, 2026, GitHub Copilot moved every plan to usage-based billing. Instead of counting a fixed number of "premium requests," each paid plan now includes a monthly allotment of GitHub AI Credits, and advanced usage draws against that balance based on the tokens each model consumes.

• Code completions and Next Edit suggestions stay unlimited on every plan, including Free — they do not consume AI Credits
• AI Credits cover the metered features — Chat, Edits, agent mode, the coding agent, and code review — and are spent based on each model's token rates
• Pro+ unlocks all models including Claude Opus 4.7 and OpenAI o3
• Overage: paid plans can buy additional usage beyond their included credits; 1 AI Credit equals $0.01
• Transition credit boost (through August 31, 2026): to ease the change, existing Business plans receive an extra $30 per user per month and Enterprise plans an extra $70 per user per month in AI Credits while teams measure real-world usage, after which allocations return to standard plan levels

GitHub also reshaped its Individual plans around the rollout — temporarily pausing new self-serve Pro and Pro+ sign-ups and tightening usage limits based on account history and verification — and introduced Copilot Max, an upgrade for Student, Pro, and Pro+ subscribers that adds higher included usage and higher spending limits for intensive agentic workflows.

Internal Use at Microsoft

In May 2026, Microsoft disclosed that it is ending an internal employee pilot of Anthropic's Claude Code on June 30 and redirecting those users to Copilot CLI. The original pilot was meant to expose project managers and designers to AI coding for the first time, but ran far over its token budget — one Microsoft employee reportedly burned through a monthly allocation in just over a week, and the program consumed a substantial share of Microsoft's full-year internal AI budget in only a few months. The decision was framed as procurement economics rather than a security or contractual dispute, and surfaced one uncomfortable signal for Copilot's positioning: when offered a free internal choice between Copilot and Claude Code, Microsoft's own developers preferred Claude for autonomous multi-file work. Microsoft is responding with a consolidated push on Copilot CLI as the single sanctioned internal coding agent, which provides Copilot one of the largest captive-developer footprints of any AI coding tool — north of 200,000 engineers — and demonstrates Microsoft's commitment to keeping its own platform competitive in agentic developer workflows. For external buyers, the takeaway is twofold: Copilot is the cheapest-by-far option for token-heavy organizations, and Microsoft is treating internal Copilot adoption as a strategic priority that will likely accelerate the product's iteration pace through 2026.

Core Capabilities

Code Completions

Copilot's original feature: as you type, it suggests single lines, entire functions, boilerplate code, and documentation inline. Accept with Tab, reject by continuing to type. Completions are context-aware — Copilot reads your open files, imports, and comments to generate relevant suggestions.

Copilot Chat

A conversational AI interface available in your IDE sidebar, on GitHub.com, in GitHub Mobile, and in Windows Terminal. Ask coding questions, explain code, generate tests, debug errors, or discuss architecture. Chat has full access to your workspace context — open files, terminal output, and project structure.

Copilot Edits

Multi-file editing from a single prompt. Describe a change in natural language, and Copilot proposes edits across multiple files simultaneously. Review the diff, accept or reject individual changes, and iterate. Available in VS Code, Visual Studio, and JetBrains.

Agent Mode

An autonomous in-IDE agent that goes beyond single edits. Give it a goal, and agent mode determines which files to read and modify, suggests terminal commands to run, recognizes errors in build or test output, and automatically attempts fixes. It operates in a loop — plan, edit, verify — until the task is complete.

Next Edit Suggestions

Copilot predicts where you will edit next based on your recent changes and automatically positions a suggestion at that location. Press Tab to accept. This feature anticipates multi-location edits — if you rename a variable in one place, it suggests the same rename everywhere else.

Copilot Coding Agent (Async)

The most distinctive feature in Copilot's lineup — generally available since September 2025, rebuilt from the earlier Copilot Workspace. Assign a GitHub Issue to Copilot, and it spins up a secure cloud environment via GitHub Actions, explores your codebase, implements the change, runs your tests and linters, and opens a draft pull request with full session logs. It works in the background — you do not need your IDE open.

Best suited for low-to-medium complexity tasks: adding features, fixing bugs, extending test coverage, updating documentation, and refactoring. The coding agent can also use MCP servers for external context. As of March 2026, improved session visibility shows setup steps (repo cloning, agent firewall activation) in real-time.

Mission Control, Spaces, and Custom Agents

Mission Control is a multi-task dashboard for managing multiple coding agents simultaneously — see all active agents, their progress, and results in one view.

Copilot Spaces are persistent context containers — save project knowledge, frequently used files, and documentation that persists across sessions. Agents working in a Space immediately have the relevant context.

Custom Agents can be defined in .github/agents/ — specialized agents for project-specific tasks (database migrations, deployment workflows, code review checklists). These agents are triggered by issues, PRs, or manual invocation.

Copilot CLI

Generally available since February 2026, Copilot CLI brings full agentic capabilities to the terminal — planning, building, reviewing, and remembering across sessions, all without leaving the command line. It has grown from a terminal assistant into a complete agentic development environment with custom agents, sub-agents, and auto-approve support for MCP.

The GitHub Copilot App

Announced at Build 2026, the GitHub Copilot app is a dedicated desktop application that GitHub describes as "a control center for agent-native development." It reflects how coding work changes once autonomous agents do much of the typing: rather than living inside a single editor window, developers increasingly direct several agents at once and need one place to steer, inspect, and merge their work.

The app's defining idea is parallel agents in isolation. Every agent session runs in its own git worktree — a real, isolated copy of your branch — so multiple agents can work at the same time without overwriting each other's changes. A unified "My Work" dashboard consolidates active agent sessions, GitHub Issues, pull requests, and background automations into a single view, removing the constant switching between terminals, IDEs, and browser tabs.

Two features stand out:

• Agent Merge carries a pull request through to completion — it monitors continuous-integration checks, tracks required reviewers, addresses failing checks, and merges once your conditions are met. You decide how far Copilot should go, from "drive the checks back to green" to "merge when everything passes."
• Canvases are shared, bidirectional work surfaces for humans and agents — a plan, a pull request, a browser session, a terminal, or a dashboard — places where an agent's intent becomes visible work you can steer and verify.

The app launched as a technical preview spanning Windows, macOS, and Linux, available to Copilot Business, Enterprise, Pro, and Pro+ subscribers.

Supported IDEs

Copilot integrates with VS Code, JetBrains, Visual Studio, Xcode, Vim and Neovim, and Eclipse. Code completions are available everywhere. The full chat, edits, and Agent Mode experience is currently only available in VS Code and JetBrains, with Visual Studio offering chat and edits but not yet Agent Mode. Xcode and Eclipse support MCP for connecting external tools.

Copilot Chat is also available on GitHub.com, GitHub Mobile, and Windows Terminal.

AI Models

Copilot offers multi-model selection — choose a specific model or use Auto mode:

• OpenAI: GPT-4.1, o3, Codex models
• Anthropic: Claude Sonnet 4.6, Claude Opus 4.7, Claude Haiku 4.5
• Google: Gemini 3.1 Pro
• xAI: Grok Code Fast 1

Pro+ subscribers get access to all models. Free and Pro tiers have a curated subset with automatic model selection.

MCP Support

GitHub Copilot supports the Model Context Protocol for connecting to external tools and data sources:

• Works in agent mode and the coding agent — Copilot can query databases, fetch documentation, or call APIs mid-task
• Supports both local and remote MCP servers
• Enterprise admins can enable or disable MCP via organization policy
• Available in VS Code, JetBrains, Eclipse, and Xcode

Strengths

• GitHub platform integration: Issues, PRs, Actions, code review, and the coding agent are all native — no third-party glue needed
• Multi-model choice: Widest model selection across four AI providers; switch models per conversation
• Async coding agent: Unique ability to work on Issues in the background via GitHub Actions — no other tool offers this
• Enterprise compliance: Microsoft-backed with policy controls, audit logs, IP indemnity, and content exclusion rules
• Distribution: Available in every major IDE; largest installed base of any AI coding tool
• Free tier: 2,000 completions/month is a strong entry point for trying AI-assisted coding

Limitations & Considerations

• Autocomplete quality: Competitors like Cursor and Claude Code offer deeper context understanding for complex multi-file edits
• Competitive landscape shift (April 2026): Microsoft reportedly explored acquiring Cursor before SpaceX secured a $60 billion option to acquire the AI-native IDE — Copilot now competes with a Cursor that sits inside SpaceX's AI stack alongside xAI and Grok
• Agent mode maturity: In-IDE agent mode is less capable than Claude Code for complex, multi-step autonomous tasks
• Enterprise cost: Enterprise plan requires GitHub Enterprise Cloud ($21/user/month additional), raising total cost significantly
• Usage-based costs: since the June 2026 shift to AI Credits, heavy chat and agent-mode use on lower tiers can exhaust the monthly credit allotment and trigger overage charges — token-hungry models burn credits fastest. The change also removed the previous unlimited fallback model, so there is no longer a zero-cost default to fall back on once credits run out — a point of friction in the developer backlash that followed the June 1 rollout
• Model availability: Best models (Claude Opus, o3) are restricted to Pro+ — lower tiers get a more limited selection

Best Use Cases

When to choose alternatives:

• Complex autonomous multi-file tasks → Claude Code (deeper reasoning, terminal-native, MCP-first)
• AI-native IDE experience → Cursor (purpose-built editor with project-wide context)
• Google Cloud ecosystem → Gemini CLI (native GCP, Firebase, Vertex AI integration)
• Privacy-first local inference → Local models via Ollama (no code leaves your machine)

Getting Started

1. Sign up at github.com/features/copilot — the Free plan requires no credit card
2. Install the GitHub Copilot extension in VS Code or your preferred IDE
3. Start typing code — inline completions appear automatically; press Tab to accept
4. Open Copilot Chat in the sidebar to ask questions about your code or request changes
5. Try Copilot Edits: select multiple files, describe a change, and review the proposed diff
6. Enable agent mode for autonomous multi-step tasks within your IDE
7. Assign a GitHub Issue to Copilot to try the async coding agent

Key Takeaways

• GitHub Copilot has evolved from an autocomplete tool into a full AI coding platform — completions, chat, edits, agent mode, and an async coding agent that works on Issues independently
• The multi-model architecture lets you choose from OpenAI, Anthropic, Google, and xAI models based on your task
• The async coding agent is unique to Copilot — assign an Issue and get a draft PR without opening your IDE
• The new GitHub Copilot desktop app (Build 2026, technical preview) is an agent-native control center: it runs parallel agents in isolated git worktrees, consolidates work in a "My Work" dashboard, and uses Agent Merge to shepherd pull requests through checks and review
• As of June 2026 all plans use usage-based billing with monthly GitHub AI Credits (1 credit = $0.01); code completions stay unlimited and free, while chat and agent features draw down credits by model token usage — Pro+ ($39/month) includes $39 in credits and unlocks all models. Amid a developer backlash, GitHub gave existing Business and Enterprise customers a temporary credit boost through August 31, 2026, paused new self-serve Individual sign-ups, and added a higher-usage Copilot Max upgrade
• Best suited for teams already on GitHub who want deep platform integration; for complex autonomous coding, consider pairing with Claude Code