What GitHub Copilot Is: From Completion to Agent
Copilot started as a completion tool and has expanded into three main layers:
- Inline completion: suggestions appear as you type inside the editor, with the lowest friction.
- IDE chat / agent mode: natural-language chat and multi-step tasks, including cross-file edits and automated handling.
- GitHub workflow: PR review, issue → PR, and cloud agent tasks.
Supported editors: VS Code, Visual Studio, JetBrains IDEs such as IntelliJ and PyCharm, Neovim, and Xcode. The most complete experience is still in VS Code.
The 2026 Pricing and Billing Changes Come First
GitHub’s official docs note:
- Usage-based billing starts on 2026-06-01: plans that used to be measured by premium requests / completions will move to usage-based billing.
- New Pro / Pro+ / student self-service signup has been paused since 2026-04-20: the launch timing depends on GitHub’s announcements.
If you cite any specific quota in a 2026 article, attach a date to it. After 2026-06-01, the quota table is going to change substantially.
Free, Pro, Pro+, Business, and Enterprise Compared
| Plan | Monthly price | premium requests / month | Main points |
|---|---|---|---|
| Free | $0 | 50 | inline completion + basic chat |
| Pro | $10 | 300 | full personal plan |
| Pro+ | $39 | 1,500 | heavy personal use / lots of Agent mode runs |
| Business | $19 / user | 300 / user | team management + policy |
| Enterprise | $39 / user | 1,000 / user | enterprise controls + full package |
This table only represents the state before the 2026-06-01 billing change. After that, compare it with the latest official pricing page.
Workflows Where Copilot Fits
- The team already runs GitHub PR review and issue tracking.
- You need organization policy, compliance, or centralized license management through Business / Enterprise.
- You are an individual developer who does not want to switch editors and wants AI inside VS Code.
- You maintain OSS and can use the free maintainer eligibility.
Workflows Where Copilot Does Not Fit
- If you mostly live in Terminal and need repo-wide tasks → Claude Code fits better.
- You want an AI-native IDE experience → Cursor has deeper integration.
- You are building your own AI coding pipeline with custom prompts → use the Codex API directly.
- If OpenAI model cost is a real concern → the token billing route through Codex can be easier to control.
Copilot vs Cursor vs Claude Code vs Codex
| Dimension | Copilot | Cursor | Claude Code | Codex |
|---|---|---|---|---|
| Entry point | editor plugin + GitHub | AI-native IDE | Terminal | OpenAI API |
| Main strength | GitHub workflow integration | IDE multi-file Composer | repo-wide / multi-agent | custom pipeline |
| Billing | subscription / moving to usage-based | subscription + Agent usage | subscription / API | token billing |
| Learning curve | half a day | half a day | 1-2 weeks | depends on the use case |
Agent Mode Notes
Once Agent mode is on, Copilot can edit across files and handle PR tasks automatically. In practice, there are a few hard rules I would not skip:
- Branch policy / required reviewers must be set first: do not let an agent push directly to main.
- Check the audit log: regularly inspect what the agent changed and whether it followed policy.
- Run CI / tests before the PR: the agent’s changes still need CI as a guardrail.
- Do not hand out external-action permissions: a GitHub token is not a license for the agent to hit your other internal APIs.
Conclusion
Copilot is still the low-friction entry point inside the GitHub ecosystem. The real subscription question is not “Is Copilot strong?” but “Does my workflow actually live in GitHub?” If 90% of your day is PR / issue / review work, Copilot is the default answer. If 90% of your day is long-running Terminal tasks, Claude Code is more direct. If you are building OpenAI-native integrations, the Codex API route usually takes less effort.