GitHub Copilot

GitHub Copilot spent 2026 doing two things at once: expanding its capability surface into agent mode, cloud agents, and a general-availability CLI, and absorbing a run of trust-eroding headlines that its dedicated competitors didn't have to deal with. Both are true, and a fair review has to hold them together rather than picking one. What it actually is: an AI pair-programming layer that plugs into editors people already use — VS Code, Visual Studio, JetBrains IDEs, Neovim, Vim, Azure Data Studio, and Eclipse — rather than a standalone AI-first IDE like Cursor or Windsurf. That architectural choice is the whole story of who should buy it. Copilot's job is to bolt AI onto your existing setup with as little disruption as possible; it is not trying to replace your editor. **Who it's actually good for.** Anyone already inside the GitHub workflow — repos, pull requests, Actions, Issues — gets a coherent experience: Copilot's chat, code review, and coding-agent features are wired directly into the same PR and issue surface the team already uses. For a developer or small team that wants competent inline completions and occasional agentic help without switching editors or paying a premium, Copilot is the low-friction option. It also supports more IDEs than most competitors, which matters for teams that aren't uniformly on VS Code. **Strengths, specifically.** Price is a real advantage: Pro is $10/month against Cursor Pro's $20, and the gap widens at the team tier ($19/seat for Copilot Business versus $40/seat for Cursor Business, per independent comparison coverage). Model choice has broadened considerably — the current lineup spans multiple providers rather than being locked to one, and Pro and above let a user pick which model handles a given task. Business and Enterprise plans bundle IP indemnity (GitHub covers legal defense if an unmodified, filter-passed suggestion triggers an IP claim), organization-level policy controls, and audit logs — features individual plans don't get. The GA of Copilot CLI in February 2026 extended the same assistant to the terminal, which several outlets covered as closing a real gap against terminal-first competitors. **What's actually wrong with this.** Three things, and none of them are cosmetic. First, trust took a real hit in March 2026 when GitHub's own VP of developer relations confirmed that Copilot had been inserting promotional "tips" — plugging Raycast, Slack, Teams, and other integrations — into pull requests, affecting a reported 1.5 million PRs before the feature was disabled. The mechanism (a hidden HTML template marker preceding the ad text) pointed to a deliberate templating decision rather than a model hallucination, which is arguably worse: it means promotional content was designed into the coding-agent output path and shipped without developers being told. Second, GitHub changed its default data-handling policy for individual users. As of April 24, 2026, interaction data — inputs, outputs, code snippets, and context — from Free, Pro, and Pro+ accounts is used to train GitHub's models by default, unless the user manually opts out in account settings. Business and Enterprise plans are excluded from this by default; individual-tier users are not. This is a meaningful reversal from the posture GitHub had been signaling, and it's opt-out rather than opt-in — the burden is on the user to notice and act. Third, the pricing and quota model changed shape on June 1, 2026: the old premium-request-unit system was replaced with GitHub AI Credits tied to token consumption. Code completions and Next Edit Suggestions stay unlimited on paid plans, but chat, agent mode, and code review now draw down a monthly credit allowance ($15 on Pro, $70 on Pro+, $200 on Max), with overage priced at roughly $0.01 per credit or about $0.04 per premium-request-equivalent once exhausted. Headline prices didn't move, but several developers covered in trade press described the change as "pay the same, get less" — worth budgeting around rather than assuming the $10 sticker price is the whole story for anyone using agent features regularly. On top of that, independent comparison coverage through 2026 consistently places Copilot behind Cursor, Claude Code, and similar tools on deep multi-file agentic work and repo-scale context handling — Copilot's own comparison writeups from third parties describe it as the lower-friction, lower-ceiling choice, not the more capable one. Community discussion also tracks a perceived quality dip since late 2025 tied to repeated underlying-model transitions, though this is pattern-level user sentiment rather than something independently reproduced here. **Pricing reality.** Five tiers: Free ($0, 2,000 completions/month, 50 chat requests, limited models), Pro ($10/month, unlimited completions, $15 in monthly AI credits, cloud agent and code review), Pro+ ($39/month, $70 in credits, access to higher-tier models including Opus), Max ($100/month, $200 in credits), and organization plans — Business at $19/user/month and Enterprise at $39/user/month, both excluding individual users' data from training and adding indemnity and admin controls. The free tier is genuinely usable for light completion work but not for agent-heavy workflows. **Verdict.** Copilot remains the sensible default for developers who want cheap, broadly-compatible AI assistance without leaving their existing editor or GitHub workflow, but it is no longer the obvious best-in-class choice for anyone doing serious agentic, multi-file work — that ground has shifted to more expensive, more capable competitors — and 2026's trust missteps (ad injection into PRs, default-on data training for individual users) mean the product now asks for more scrutiny of its settings than it used to. ## Quick take **Verdict:** GitHub Copilot is the cheapest, most editor-agnostic way to get competent AI coding help inside a workflow you already have, but it trails dedicated agentic tools on deep autonomous work and had a rough enough 2026 on trust that its default settings are worth checking before you rely on it. **Choose it if:** - You're already living in VS Code, Visual Studio, JetBrains, Neovim, or Eclipse and want AI help without switching editors - Price and broad IDE coverage matter more than the deepest possible agentic capability - Your org already pays for GitHub Enterprise and wants IP indemnity, audit logs, and admin controls bundled in at the Business/Enterprise tier - You want a single vendor for source control, CI, PRs, and AI assistance **Skip it if:** - You want the strongest available autonomous multi-file agent performance — independent 2026 comparisons consistently rate Cursor and Claude Code higher here - You're on an individual plan and privacy-sensitive — GitHub trains on Free/Pro/Pro+ interaction data by default as of April 24, 2026 unless you manually opt out - You need pricing stability — the June 2026 shift from premium-request units to token-metered AI Credits changed how far a subscription actually goes - The March 2026 incident where Copilot inserted promotional content into pull requests is a dealbreaker for you on principle **Watch out:** GitHub began training on Free, Pro, and Pro+ users' interaction data by default on April 24, 2026 — opting out requires a manual settings change, and only Business/Enterprise plans are excluded automatically. Separately, the June 1, 2026 move to usage-based AI Credits means chat, agent, and code-review usage now draws down a monthly allowance even though headline plan prices stayed the same.