GitHub Copilot — Multi-IDE AI Pair Programmer With Agent Mode, PR Review & 20+ Model Picker (Microsoft/GitHub, Active, Not Shut Down)

Not shut down. GitHub Copilot launched 2021 (GitHub + OpenAI), now Microsoft/GitHub property, active 2026-08, no acquire/rebrand/shutdown. June 1 2026 moved from premium-request counting to GitHub AI Credits (1 credit = $0.01, token-based, base+flex allotment, no rollover). Category ai-productivity-data; pipeline_stage none. Killer differentiator = deepest platform reach of any AI coder (VS Code / JetBrains / Visual Studio / Neovim / Xcode / Eclipse / Zed / GitHub web / Mobile / CLI / Windows Terminal / Raycast) + one auth/billing across completions+chat+agent+PR-review+Spark + model picker spanning GPT-5.6/Claude Opus 4.8/Gemini 3.x/Grok 4.6/Kimi/MAI + Business/Enterprise pooled credits + code-review agent inline on PRs; not Cursor (VS Code fork only, deeper editor-native agent), not Claude Code (strongest autonomous agent, no JetBrains/Xcode), not Tabnine (privacy/airgap first, weaker gen), not Codeium/Windsurf (cheaper, smaller enterprise governance). 2026 trap: June 2026 credit migration burned heavy-agent users (allowance drained in days vs old PRUs), completions unlimited but chat/agent/review/CLI all draw credits, frontier models (Opus 4.8/GPT-5.6 Sol) 10×+ cheaper model per token, code-review also eats GitHub Actions minutes, individual plans train-on-by-default (Business/Ent contractual opt-out), Free 2,000 completions only + auto-model-no-pick, G2 ~4.5 but satisfaction surveys show 18% dissatisfied vs Cursor 6%.

1. One-line positioning

Copilot = GitHub-native AI engineering layer: type → inline completion + Next Edit Suggestions (unlimited, zero credits) → Chat in IDE/GitHub/CLI → Agent Mode plans+makes+tests across files → assign a GitHub issue to Copilot cloud agent, it opens a self-reviewed PR → PR diff gets inline review comments (bugs/edge-cases/security) → model picker swaps GPT-5.6 ↔ Claude Opus 4.8 ↔ Gemini 3.7 Flash ↔ Grok 4.6 per task → Spaces ground org context → MCP + third-party agents (Claude Code, Codex) on Pro+. Fits ai-productivity-data; pipeline_stage none (type→complete→chat→agent→review→merge). For teams already on GitHub+VS Code/JetBrains who want one governed seat, it’s the lowest-friction default; for agent-maximalists on complex multi-repo refactors, Cursor+Claude Code edge out.

2. Core features

Layer	Capability	Notes
Completions	Inline + Next Edit Suggestions, 40+ langs, unlimited on all paid plans, zero credits	Free = 2,000 completions/mo cap
Chat	IDE inline / GitHub.com / Mobile / CLI / Windows Terminal / slash commands / skills	Free = limited chat, auto-model only
Agent Mode	Multi-file plan-edit-test loop in IDE; cloud agent (issue→branch→PR) on Pro+	Pro+ can delegate to Claude Code / Codex
Code Review	PR diff analysis, inline comments, catches unhandled promises/null checks, GA Mar 2026	Consumes AI credits + GitHub Actions minutes
Model picker	GPT-5.6 Sol/Luna/Terra, Claude Opus 4.8/5/Sonnet 4.6, Gemini 3.7 Flash/3.5/3.1 Pro, Grok 4.6, Kimi K3, MAI-Code-1.1, o4-mini	Free/Student = auto-select only, no pick
Workspace / Spark	Browser agentic dev env (issue→plan→PR), GitHub Spark (public preview)	Pro+ and Enterprise
Spaces	KB replacement (Nov 2025), ground chat/agent in repo docs/specs	Ent-first, Pro+ gets it
Integrations	100+ via MCP, GitHub Actions, Linear/MS Teams/Azure Boards hooks, CLI, Copilot App	No native Figma/Notion
Governance	SSO/SAML, audit logs, IP indemnity (Biz+), content exclusions, block public-code match, BYOK (Ent), budget caps + pooled credits (Biz/Ent)	SOC 2; Biz/Ent no-train contractual

3. Pricing (2026-08, USD, AI Credits 1 = $0.01)

Tier	Monthly	AI credits/mo (base+flex)	Key gates
Free	$0	limited (not published)	2,000 completions, auto-model chat only, no agent/workspace/Spaces, no credit picker
Student	$0 (verified)	limited	Like Free, unlimited completions, no third-party agents
Pro	$10	1,500 (1,000+500)	Unlimited completions, agent mode, cloud agent, 3rd-party agents, model pick, $15 value on $10
Pro+	$39	7,000 (3,900+3,100)	Premium models (Opus 4.8 etc), 4×+ Pro usage, audit logs, Spaces
Max	$100	20,000 (10,000+10,000)	Priority new models, sustained agent herds
Business	$19/seat	1,900/seat pooled	Org policy, budget caps, IP indemnity, no Pro+ Max individual headroom
Enterprise	$39/seat	3,900/seat pooled (+ promo 7,000 through Aug 2026)	SAML, BYOK, codebase indexing, Spaces, priority models

pricing_type = freemium-per-seat-usage-credit-metered; starting_price = $0; free_tier = "yes; 2,000 completions/mo, limited chat (auto model), no agent/workspace/model-pick/Spaces, community support"; pricing_note = "jun-1-2026-premium-requests-to-ai-credits-1credit-0.01-token-based-no-rollover; completions-next-edit-zero-credit-always; chat-agent-review-cli-spark-consume-credits; frontier-models-opus-4.8-gpt-5.6-sol-10x-cheaper-model-per-token; code-review-also-bills-actions-minutes; pro-1500-credits-proplus-7000-max-20000-biz-1900-pooled-ent-3900-pooled; jun-aug-2026-promo-biz-3000-ent-7000-existing-only; individual-plans-train-on-default-biz-ent-contract-opt-out; new-annual-individual-paused-monthly-only"

4. Access Type

access_type: ide-plus-github-web-plus-cli-plus-mobile

access_display: 🖥️ VS Code / JetBrains / Visual Studio / Neovim / Xcode / Eclipse / Zed · GitHub.com PR/issue/chat · 📱 iOS/Android · 💻 Copilot CLI + Windows Terminal · 🔌 MCP + 3rd-party agents (Pro+) · no standalone desktop app

5. Reviews (2026)

Source	Read
G2	4.5/5 (370+ rev) — praise: kills boilerplate, tab-to-accept flow, PR review catches real bugs; gripes: confident-wrong imports, loses thread in 1k-line files, June 2026 credit drain
Capterra	~4.4/5 — “most convenient because it’s already in the editor+GitHub”, caution on over-reliance
Dev survey (400 ppl Feb 2026)	Satisfaction 52% (vs Cursor 78%, Claude Code 84%), dissatisfaction 18%; market share 24% (down from 28% 2025)
Community threads	Post-June-1 2026: “allowance that lasted a month gone in 4 days on cloud agent” → some migrated to Claude Code/Cursor
Enterprise buyers	Gartner/Peer Insights: governance+audit+contractual no-train beats technically stronger rivals; productivity real but needs human review

Praise: completions still best-in-class tab flow, one auth across editor+GitHub+CLI, PR review agent genuinely catches null/unhandled-promise bugs, JetBrains/Xcode coverage no rival matches, Business $19 pooled credits cheapest team seat in category, model picker avoids single-vendor lock.

Gripes: credit metering opaque post-June 2026, frontier-model agent runs burn $70+ fast, context drops on 10+ file architectural refs (Cursor Composer/Claude Code sharper), Free tier auto-model-only feels crippled, individual train-on-default spooks cautious devs, no live chat support for individuals, code-review double-bills credits+Actions minutes, satisfaction lagging Cursor/Claude Code in 2026 polls.

6. Best for / Not for

Best for	Not for
Teams already on GitHub + VS Code/JetBrains wanting one governed seat	Agent-maximalists doing 10+ file cross-repo refactors daily (Cursor+Claude Code sharper)
Enterprise with audit/SSO/IP-indemnity/SOC2 needs	Privacy-paranoid orgs banning cloud egress (Tabnine airgap / Continue+local model)
Java/Kotlin/Scala backend on IntelliJ (no Cursor equiv)	VS Code-only solo who wants editor-native Composer (Cursor $20)
Maintainers wanting PR review agent on every diff	Lightweight script writers who’ll never exceed Free 2,000 completions
Pro devs using mixed models per task (cheap Luna for boilerplate, Opus for debug)	People who won’t watch credit dashboard (surprise overage)

7. Competitors

Tool	Lane	Starts
Cursor	VS Code fork, Composer 2 agent, deepest editor-native	$20/mo
Claude Code	Autonomous agent, strongest multi-file rep, no JetBrains	~$40–100 usage / Max $100
Codeium / Windsurf	Cheaper ($15), fast codebase chat, lighter governance	$15/mo
Tabnine	Privacy-first, airgap, enterprise on-prem	custom
JetBrains AI Assistant	IntelliJ-native, no VS Code	$10/mo w/ IDE
Replit AI	Browser IDE + agent	$15/mo
Continue.dev + local LLM	Zero egress, OSS	$0