GitHub Copilot

My hands-on review of GitHub Copilot

GitHub Copilot is no longer limited to suggesting the next line of code. Its current product is much closer to an AI development environment layered across GitHub, IDEs, CLI, and agent workflows.

The basic autocomplete experience remains useful. While typing code, Copilot can suggest functions, expressions, boilerplate, and larger blocks of implementation. This is particularly useful for repetitive coding tasks.

The bigger change is Agent Mode. Instead of asking Copilot to generate one function at a time, a developer can give it a broader task. Copilot can inspect the project, modify multiple files, run commands, and continue working until the requested task is completed or requires developer input.

Copilot cloud agent takes this further by operating through GitHub. It can research a repository, create an implementation plan, make iterative changes on a branch, allow the developer to inspect the diff, and create a pull request.

Another important feature is model choice. Current Copilot plans provide access to a range of models, with availability depending on the plan and feature. GitHub has also moved Copilot usage to an AI Credit system where model usage and token consumption determine credit consumption.

MCP support makes Copilot more extensible because developers can connect supported MCP servers and allow agents to interact with additional tools and services.

For professional development, the combination of code completion + chat + agents + GitHub + CLI + code review makes Copilot significantly broader than a traditional autocomplete assistant.

How GitHub Copilot actually works

GitHub Copilot operates through several connected development experiences.

The simplest workflow is inline completion. As a developer writes code, Copilot analyzes the surrounding context and proposes code suggestions.

The developer can also open Copilot Chat to ask questions about code, errors, architecture, APIs, or a repository.

For larger tasks, Agent Mode can be used. The developer gives Copilot a goal such as:

"Add authentication to this application and update the affected routes, components, tests, and documentation."

Copilot can inspect the project, determine which files need modification, make changes, and run commands as part of the workflow. Developers can also use plan mode before allowing changes, or delegate specific subtasks to subagents.

Copilot cloud agent provides another workflow:

Repository → Research → Plan → Implement → Iterate → Review Diff → Pull Request

The cloud agent can work on a repository hosted on GitHub and perform multi-step development tasks on a branch.

Copilot also works from the terminal through Copilot CLI, allowing developers to interact with the coding agent without leaving the command line.

MCP extends Copilot's capabilities by allowing compatible agents to connect to external tools and services through MCP servers. GitHub currently lists MCP support across Copilot plans.

The result is a development workflow where AI can participate in everything from individual code suggestions to larger repository-level implementation tasks.

Pricing: is it worth the money?

GitHub currently offers several Copilot plans for individuals and organizations.

Copilot Free costs $0/month and currently includes 2,000 code completions per month, limited AI usage, Copilot CLI, and access to selected models.

Copilot Pro costs $10/user/month. It adds unlimited code completion and next-edit suggestions, cloud agent, code review, model selection, third-party agents, and a monthly AI Credit allowance. GitHub currently lists 1,500 total monthly AI Credits for Pro under its current usage model.

Copilot Pro+ costs $39/user/month and provides substantially more included AI usage plus access to premium models and additional features such as audit logs. GitHub currently lists 7,000 total monthly AI Credits for this plan.

Copilot Max costs $100/user/month and is designed for high-volume agent workflows. GitHub currently lists 20,000 total monthly AI Credits for Max.

For organizations, Copilot Business is currently $19/user/month, while Copilot Enterprise is $39/user/month. Business includes centralized management, governance, pooled credits, cloud agent, code review, and broad model access. Enterprise adds higher included usage and additional enterprise capabilities.

GitHub's current billing system uses AI Credits, where 1 AI Credit = $0.01 USD. AI interactions consume credits based on the model and token usage. This means expensive models and long agent sessions can consume more credits than simple lightweight interactions.

This is important for agencies because high-volume autonomous coding can create usage costs beyond the basic subscription.

GitHub Copilot vs. the alternatives

• 1. Cursor

  AI-first code editor focused on repository-aware coding, agentic development, multi-file changes, and advanced AI workflows.

• 2. Claude Code

  Terminal-based coding agent designed to understand codebases, modify files, execute development tasks, and work through complex coding problems.

• 3. OpenAI Codex

  AI coding-agent technology for delegating software-development tasks, code generation, debugging, and repository-level work.

• 4. Windsurf

  AI-powered development environment combining code completion, conversational coding, and agentic workflows.

• 5. Amazon Q Developer

  AI coding assistant from AWS designed for software development, code generation, debugging, modernization, and AWS-focused development workflows.

• 6. Gemini Code Assist

  Google's AI coding assistant for IDEs and developer workflows, with code generation, debugging, explanations, and Google Cloud integration.

• 7. Tabnine

  AI coding assistant focused on code completion, enterprise development, privacy, customization, and developer productivity.

• 8. Codeium / Windsurf

  AI development tooling focused on code completion, chat, codebase understanding, and agentic coding workflows.

• 9. JetBrains AI Assistant

  AI development assistant integrated directly into JetBrains IDEs for coding, explanations, refactoring, documentation, and development workflows.

• 10. Replit Agent

  AI-powered development environment that can turn natural-language instructions into applications and handle significant portions of the build process.

My verdict

GitHub Copilot has developed into a broad AI coding and software-engineering platform rather than simply an autocomplete extension.

Its biggest strength is its integration with the development environment developers already use. Copilot can work inside IDEs, GitHub, and the terminal, allowing AI assistance to remain close to the actual codebase and development workflow.

The addition of Agent Mode and Copilot cloud agent changes how developers can interact with the product. Instead of asking for isolated snippets, developers can delegate larger tasks such as repository research, implementation, refactoring, testing, and pull-request preparation.

Its support for multiple models, MCP, custom instructions, third-party agents, code review, and CLI workflows makes it useful across different development styles.

The main consideration is that Copilot does not eliminate the need for engineering judgment. AI-generated code can contain bugs, incorrect assumptions, security issues, or architectural problems. Developers still need to review changes, run tests, validate dependencies, and understand what gets merged into production.

The newer AI Credit system is another consideration. Basic coding assistance can fit comfortably within a plan, while extensive agent usage and premium models can consume credits much faster.

For developers and development agencies, Copilot can function as a productivity layer across the complete software-development lifecycle—from writing code to researching repositories, implementing changes, reviewing code, and preparing pull requests.

GitHub Copilot FAQ