GitHub Copilot is GitHub’s AI coding product, and the page that introduces it now describes it as an AI coding agent. It started as a completion tool, and the pages we read show a wider set of features: chat, a cloud agent that turns issues into pull requests, code review and a command line tool. This guide covers what GitHub documents on its public pages, how the plans are priced and where the limits are.
We did not run hands-on tests, so we cannot comment on the quality of suggestions or how well the agent handles real repositories. If you want a shorter read first, see our shorter review. For choosing between coding tools, our guide on how to shortlist AI tools for developers gives a method.
How we researched this: we did not run hands-on tests for this article. It is based on the vendor’s own documentation and public sources, checked on 11 November 2026. Features and prices change, so confirm details on the vendor’s site before you buy.
Key takeaways
- GitHub’s own pages describe Copilot as an AI coding agent that works in editors, on github.com, in a command line tool and in chat apps.
- On 14 November 2026 the plans page listed Free at $0, Pro at $10, Pro+ at $39 and Max at $100 per user a month, with Business and Enterprise listed without prices.
- Completions are limited to 2,000 a month on Free, and paid plans are listed with unlimited completions and next edit suggestions that do not use AI credits.
- Chat and agent work draw on GitHub AI Credits, where one credit equals one US cent, so heavy use can matter more than the plan price.
- We could not read GitHub’s privacy or data use pages, and the plans page has one inconsistency about third party agents on Pro, so confirm details before you buy.
What GitHub Copilot is and who makes it
GitHub Copilot is a product of GitHub. The Copilot page on github.com describes it as an AI coding agent. It lists the places you can use it: VS Code, Visual Studio, JetBrains, Xcode, Vim, Neovim, Zed and Eclipse editors, a desktop app, github.com, mobile, and chat apps Slack and Teams.
The core idea on the page is that Copilot works at several levels. There are completions as you type, chat, a cloud agent that takes an issue and turns it into a pull request, a code review feature that works on local changes and on pull requests, and a command line tool. The page also describes model choice: you can let Copilot pick a model automatically, choose a model yourself, bring your own key, or use local models through tools such as Ollama or Foundry Local.
We read the features page and the plans page. We did not read the documentation that explains each feature, so the descriptions below are limited to what those two pages state.
| Item | What the vendor pages show | Source |
|---|---|---|
| Description | An AI coding agent | github.com/features/copilot |
| Editors listed | VS Code, Visual Studio, JetBrains, Xcode, Vim, Neovim, Zed, Eclipse | github.com/features/copilot |
| Other surfaces | Desktop app, github.com, mobile, Slack, Teams | github.com/features/copilot |
| Model options | Auto select, choose, bring your own key, local models | github.com/features/copilot |
| Check date | 14 November 2026 | Our check |
Who GitHub Copilot is for
The plan structure points to a few groups. Individual developers can start with the Free plan, then move to Pro for unlimited completions. Heavy users of agent features or premium models may look at Pro+ or Max, which come with larger credit allotments. Organisations look at Business and Enterprise, where the page lists no prices and mentions admin usage limits and pooled usage.
The table maps these cases to the plans. It reflects our reading of the plans page rather than anything GitHub states about its customers.
| If you are | Plans to look at | Why, according to the page |
|---|---|---|
| Trying it | Free ($0) | 2,000 completions a month, CLI, agent mode and MCP listed |
| An everyday individual | Pro ($10) | Unlimited completions, cloud agent and code review listed |
| A heavy user of premium models | Pro+ ($39) | Larger credit total and premium models including Opus listed |
| A very heavy user | Max ($100) | Largest credit total and priority access listed |
| An organisation | Business or Enterprise | No prices on the page, admin limits and pooled usage listed |
Core features
Code completions
Completions are the longest standing feature. The plans page lists 2,000 a month on Free and unlimited completions on paid plans. It also lists next edit suggestions, which on paid plans do not use AI credits.
Chat
Chat is listed as drawing on GitHub AI Credits, where one credit equals one cent. That means chat is metered, so the amount you can do depends on your plan allotment rather than a flat count.
Agent mode
The Free plan lists agent mode, along with the CLI and MCP support. We did not read how agent mode works in detail, but its presence on Free means you can look at it without paying.
Copilot cloud agent
The features page describes a cloud agent that turns an issue into a pull request. Pro lists the cloud agent. This is one of the main reasons for the agent description, and it is worth testing on a low risk repository.
Code review
Code review runs on local changes and on pull requests, according to the features page. Pro lists code review. We did not read what it checks or how often it errs, so treat it as an assistant to a human reviewer, not a replacement.
Command line tool
Copilot CLI is listed on the features page and on the Free plan. For developers who live in the terminal, that means the product is not tied to one editor.
Newer and advanced features and integrations
Model choice and bring your own key
The features page says you can use auto select, pick a model, bring your own key or run local models. The last two options matter to teams with specific vendor or data rules, though we did not read the setup steps.
Premium models on higher plans
Pro+ is listed with premium models including Opus. The Free plan lists Haiku 4.5 and GPT-5 mini. The names are as the page shows them, and the models offered may change.
Third party agent delegation
Pro+ lists third party agent delegation and audit logs. The page is inconsistent on Pro: bullet text mentions Claude Code and Codex agents while the comparison table marks delegation as not included. We could not resolve that, so ask GitHub if it matters to you.
Copilot Fusion levels
The plans page shows Fusion at Standard, High and Max levels. We did not find an explanation in what we read, so we do not describe it.
Chat app integrations
The features page lists Slack and Teams among the surfaces. We did not read details of what each integration can do.
Pricing and plans
We read GitHub’s plans page on 14 November 2026. Prices are in US dollars per user per month. GitHub may change them, so confirm on the vendor page.
The credit system is the part to understand. Each paid plan has a base credit amount and, for some, a flex allotment, and one credit equals one cent. Chat and agent work draw on credits, while completions and next edit suggestions on paid plans do not.
| Plan | Price per user per month | Credits and notes |
|---|---|---|
| Free | $0 | 2,000 completions, Haiku 4.5 and GPT-5 mini, CLI, agent mode, MCP |
| Pro | $10 | Base credits $10, total $15 with a $5 flex allotment, unlimited completions, cloud agent, code review |
| Pro+ | $39 | Total $70 in credits, premium models including Opus, audit logs, third party agent delegation |
| Max | $100 | Total $200 in credits, priority access |
| Business and Enterprise | Not listed | Admin usage limits, pooled usage |
SSO and IP indemnity are not listed on the individual plans in what we read. If you need them, look at the organisation plans, and ask GitHub for current prices. Our guide to reading an AI tool’s pricing page explains how to compare credit based pricing.
The totals shown with a flex allotment mean the figure you pay and the credits you get are not the same number. For example, Pro costs $10 and lists a total of $15 in credits. Check whether flex credits behave differently from base credits before you rely on them.
Strengths
Many places to use it
The features page lists a long set of editors, a desktop app, mobile, github.com and chat apps. If your team uses mixed tools, that breadth is a clear point in its favour.
A real free plan
Free is $0 with 2,000 completions a month and access to chat, CLI, agent mode and MCP. That is enough to form an opinion before paying.
Unlimited completions on paid plans
Paid plans list unlimited completions and next edit suggestions that do not use credits. For people who mostly want inline suggestions, that makes cost predictable.
Flexible model options
Auto select, manual choice, bring your own key and local models give teams several routes, which matters when policies restrict which models can be used.
Close to where the code lives
Because Copilot works on github.com, with issues and pull requests, the cloud agent and code review can fit into an existing review flow. That is our reading of the features page, not a tested claim.
Weaknesses and limits
Credit based metering is complex
Chat and agent work draw on credits, with base and flex amounts and different totals by plan. It is harder to predict cost than a flat price.
Organisation prices are not on the page
Business and Enterprise show no prices in what we read. Companies need to contact GitHub or check another page.
An inconsistency on third party agents
The Pro plan text and the comparison table disagree about third party agent delegation. We could not tell which is right.
Unexplained terms
Fusion levels and some other names appear without explanation on the page we read.
Individual plans lack some enterprise features
SSO and IP indemnity are not listed on individual plans. Teams with those requirements need the organisation plans.
Privacy, data handling and security
We did not read GitHub’s privacy statement, its Copilot data use terms or its trust centre pages, so we cannot say how prompts and code are stored, whether they are used for training or how long they are kept. This is a real gap in this guide.
What the plans page does say is that Pro+ lists audit logs, and that Business and Enterprise have admin usage limits. Those are controls, not data handling statements. If you work with private code, read GitHub’s data pages and have your security team review them. Our guide on protecting company data when using AI tools lists useful questions.
How GitHub Copilot compares
GitHub Copilot’s closest rivals among our tools are Cursor, Replit and Windsurf. We compare Copilot with Cursor in our Cursor and Copilot article, and we cover recent shifts in what changed in AI coding tools this quarter.
The table compares only what we read for each tool. The products measure usage differently, so do not read it as a ranking.
| Tool | Entry paid price we read | How usage is measured |
|---|---|---|
| GitHub Copilot | Pro at $10 per user a month | Unlimited completions, chat and agents use AI credits |
| Cursor | Plus at $8 a month | Usage pool shared with Grok, refreshed weekly |
| Replit | Core with $20 in monthly credits | Credits, with Agent allowances |
| Windsurf (now Devin Desktop) | Pro at $20 a month | Quotas that refresh daily and weekly |
See also our Cursor complete guide for the plan ladder of that product.
Getting started
A cautious way to evaluate GitHub Copilot:
- Open github.com/features/copilot/plans and note current prices and credit allotments.
- Start on Free and use the 2,000 completions on a non sensitive project.
- Try agent mode and the CLI on a small repository.
- Check how many credits a typical chat session consumes.
- Read GitHub’s data use pages before using it on private code.
- Move to Pro only if you reach the Free limits, and watch credit use in the first month.
Frequently asked questions
Is GitHub Copilot free?
There is a Free plan at $0 on the plans page, with 2,000 completions a month, Haiku 4.5 and GPT-5 mini, the CLI, agent mode and MCP.
How much does Copilot Pro cost?
On 14 November 2026 the plans page listed Pro at $10 per user per month, with $10 in base credits and $15 in total with a $5 flex allotment.
What is an AI credit?
The pages say one GitHub AI Credit equals $0.01. Chat and agent work draw on credits, while completions and next edit suggestions on paid plans do not.
What are Business and Enterprise prices?
The plans page did not list prices for these plans in what we read. Ask GitHub or check its organisation pricing page.
Can I use other models?
The features page says you can use auto select, choose a model, bring your own key or use local models through Ollama or Foundry Local.
Does Copilot train on my code?
We could not confirm this. We did not read the data use pages. Check them before using private code.
Is there a shorter review?
Yes, see our shorter review, and how to choose an AI tool for method.
Conclusion
GitHub’s public pages show Copilot as a broad coding product with many editors, a cloud agent, code review and a clear plan ladder from Free to Max. The most important thing to understand before buying is the credit system, since chat and agent use draw on credits while completions on paid plans do not.
Several things remain unconfirmed: the privacy terms, organisation prices, the third party agent inconsistency and the meaning of Fusion. Try the Free plan, check credit use for a month and confirm data handling with GitHub. Our guide on checking AI output for mistakes and testing before you pay help with the next step.
One more practical point is that credit based plans reward people who watch their usage. In the first month, note how many credits a typical day of chat and agent work consumes, then compare it against the allotment on your plan. If you are well below it, a cheaper plan may do. If you hit the limit regularly, compare the cost of a higher plan with the cost of extra usage. That simple habit tells you more than any headline price, and it keeps the decision grounded in your own work rather than in marketing pages.
Before you buy, it helps to write down a few questions for your own situation. Which editor do you use, and is it on the list of supported editors? Do you mostly want completions, which are unlimited on paid plans, or do you expect to lean on chat and the cloud agent, which draw on credits? Does your employer have a policy about which models or vendors are allowed, and does the bring your own key option help with that? Does your security team need audit logs, SSO or indemnity, and if so, which plan lists them? Each answer points to a different plan, and none of them can be settled from a headline price.
It is also sensible to check how the product fits into your review process. A cloud agent that opens pull requests can save effort, but every pull request still needs a human reader. The same applies to automated code review, which should be treated as an extra pair of eyes rather than a gate. We have not tested either feature, so we cannot say how often they are right, and we recommend trying them first on a repository where a mistake costs little.
Sources
- GitHub Copilot features page, read 14 November 2026, accessed 14 November 2026
- GitHub Copilot plans page, read 14 November 2026, accessed 14 November 2026
