Back to AI Radar
ProvenDeveloper ExperienceNew entryMarch 2026 New Items

Battle-tested in production. Build on it with confidence.

GitHub Copilot

The default AI coding tool for most teams, now with full agentic capabilities, CLI GA, and multi-model choice, but premium request limits shape how you actually use it.

Agentic·DevTool·LLM·Context

github.com

Our Take

What It Is

GitHub Copilot is the AI coding assistant embedded across GitHub's developer platform. It covers code completions, an interactive chat, agent mode for multi-step tasks, automated code review, and a terminal CLI agent. You can choose between Claude Opus 4.6, GPT-5.3-Codex, and Gemini 3 Pro depending on the task.

Why It Matters

Copilot's significance isn't the AI itself but where it lives. It's the only tool with native integration into GitHub's full ecosystem: issues, pull requests, code review, Actions, and now the CLI. For teams already on GitHub, nothing else offers this depth of workflow integration. The February 2026 CLI GA release extended agentic coding to the terminal, and multi-model support means you're not locked to one provider's strengths.

Key Developments

  • Mar 2026: Agent capabilities (custom agents, sub-agents, plan agent) now GA in JetBrains IDEs; MCP auto-approve support added.
  • Feb 2026: Copilot CLI reached general availability with multi-model support (Claude Opus 4.6, GPT-5.3-Codex, Gemini 3 Pro).
  • Feb 2026: Code review via CLI now available; Visual Studio 2026 integration improved with syntax-highlighted completions.
  • Dec 2025: Improved Copilot Chat and agent mode refinements in JetBrains IDEs.

What to Watch

The premium request model is the key constraint. Free gets 50/month, Pro gets 300, Pro+ gets 1,500. Agent mode chews through these fast, and the gap between free and usable is significant ($39/month for Pro+). Watch whether GitHub loosens these limits as competition from Claude Code and Gemini CLI intensifies. Also track how well the cloud-based agent (currently 90+ second cold starts) evolves for iterative work.

Strengths

  • Platform integration: Deepest integration with GitHub's ecosystem. Issues, PRs, code review, Actions, and CLI all connected in one workflow.
  • Multi-model flexibility: Choose between Claude Opus 4.6, Sonnet 4.6, GPT-5.3-Codex, and Gemini 3 Pro. No vendor lock-in at the model layer.
  • IDE coverage: Works in VS Code, JetBrains, Visual Studio, Xcode, and the terminal. Broadest IDE support of any AI coding tool.
  • Enterprise readiness: Business and Enterprise tiers with policy controls, audit logs, and IP indemnity. Proven at scale in large organisations.
  • Free tier exists: 2,000 completions and 50 premium requests/month at zero cost, enough to evaluate properly.

Considerations

  • Premium request limits: Power users hit walls on Free (50/month) and Pro (300/month). Agent mode and chat consume requests fast, pushing you toward the $39/month Pro+ tier.
  • Cloud agent cold starts: The web-based coding agent takes 90+ seconds to spin up. If it dies mid-task, you're waiting through another cold boot.
  • Context window management: Despite file access, Copilot analyses roughly 10% of project code and fills gaps with assumptions. Multi-file reasoning still trips up on complex business logic.
  • Rate limit exhaustion: Extended agent mode sessions degrade toward the end of long conversations as limits are consumed unevenly.

Resources

More in Developer Experience

GitHub Copilot· Gemini CLI· LiteLLM· Coding Agents· Cursor· Google Antigravity· OpenRouter· Windsurf· Xcode Agentic Coding· Claude Code· OpenAI Codex· Prompt Caching

Back to AI Radar