GitHub Copilot review

GitHub Copilot is a hosted AI coding agent and SaaS product integrated with GitHub. It is not a library or a server that you install in your own infrastructure. Its workflow spans GitHub issues, pull requests, repository history, and automations, with additional access through a desktop app, CLI, editor integrations, web, and mobile. The agent can start from an issue or pull request, work in a separate session, and return changes for review.

In a software stack, Copilot sits alongside your source-control and collaboration tools, not in place of them. It uses repository context to help explore code, plan work, automate tasks, review changes, and address merge conflicts. Developers can set repository instructions and skills, connect services through Copilot Connectors, and find MCP servers, skills, and plugins. GitHub also lists support for cloud model providers using your own API key and local models through Ollama and Foundry Local. The product page does not present Copilot as a standalone API service.

GitHub presents Copilot through user-facing integration surfaces rather than a documented API-first workflow: the Copilot CLI, desktop app, GitHub.com, GitHub Mobile, and editor integrations. Listed IDEs include Visual Studio Code, Visual Studio, JetBrains IDEs, XCode, Vim, Neovim, Zed, and Eclipse. Copilot is also available in Slack and Teams. The homepage says it is trained on languages found in public repositories; it does not provide a language-by-language support matrix there.

For extending workflows, GitHub lists repository instructions and skills, Copilot Connectors, MCP servers, and plugins. The homepage does not name a Copilot SDK or API protocol, so confirm SDK support, authentication, and endpoint availability in GitHub’s documentation before planning a custom integration. The listed model options include supported cloud providers with your own API key and local models through Ollama and Foundry Local. Cross-surface session continuity is useful when work starts in one interface and continues in another, but teams should verify which integrations support the specific agent features they need.

GitHub prices Copilot per user per month. Free costs $0 and includes 2,000 completions per month, Copilot CLI, and community support; it does not include the paid plans’ unlimited completions or cloud and code-review access. Pro is $10 per user per month and includes $15 in monthly total AI credits. Pro+ costs $39 and includes $70 in credits, while Max costs $100 and includes $200. The pricing page lists additional monthly flex allotments of $5 for Pro, $31 for Pro+, and $100 for Max.

Those included credits make agent-heavy usage a variable to track: additional GitHub AI Credits can be purchased beyond the plan allowance on Pro, Pro+, and Max. The page does not publish a per-request rate or a fixed number of agent tasks each credit amount buys. Business is mentioned in the plan comparison, but its price is not published on this page. Enterprise is included with GitHub Enterprise per the page banner; contact sales for pricing. No annual discount is listed.

GitHub’s Copilot product page does not publish a software license or source-availability terms for the product. Copilot is presented as a hosted GitHub service, not a server that customers can self-host. It can connect to supported cloud model providers with your own API key or use local models through Ollama and Foundry Local, but those options do not make Copilot itself self-hosted.

The product draws on GitHub repositories, issues, pull requests, and history, so workflows built around its sessions, instructions, and automations may be tied to GitHub. GitHub does not publish data-export details on the product page. Before adopting it across a team, ask GitHub what Copilot data can be exported, how it is retained, and what terms apply to code and prompts. Moving to another tool may mean rebuilding integrations and instructions; repository code remains in your existing source-control system.

Before making Copilot part of a production workflow, check credit usage, the current availability of your chosen models, and the behavior when a session reaches its allowance. The pricing page lists monthly credits and optional additional AI Credits, but does not publish a fixed task count or per-request rate. For security-sensitive repositories, review how repository context, code review, instructions, connectors, and audit logs fit your policies. Pro+ lists audit logs; Enterprise lists SAML SSO, usage metrics, and IP indemnity as not included in lower tiers.

Reddit discussions show mixed user sentiment. Some commenters describe Copilot as good value, while others report frustration about model-specific limits, changes to model access, and communication around plan changes. Those are user reports, not published service guarantees. The product page does not provide rate-limit figures or reliability commitments. Teams should confirm current terms with GitHub, test fallback workflows, and avoid assuming a selected model will remain available on unchanged terms.