GitHub Copilot Complete Guide¶
A comprehensive guide series that systematically covers everything from instruction design to agent operations in GitHub Copilot. This includes hybrid operations with Claude Code for enterprise teams, as well as unified multi-tool management through Agent Skills and AGENTS.md.
🔀 Which Article Should You Read First?¶
graph TD
Q1{"Have you set up<br>instruction files before?"}
Q2{"Mixed technology<br>stacks?"}
Q3{"Want to hand off procedures<br>for autonomous execution?"}
Q3b{"Need deterministic control<br>over tool execution?"}
Q4{"Using AI tools<br>other than Copilot?"}
Q5{"Running out of<br>AI Credits or legacy<br>premium requests?"}
A1["Custom Instructions<br>Complete Guide"]
A2["applyTo<br>Pattern Guide"]
A3["Agent Skills<br>Guide"]
A3b["Hooks<br>Complete Guide"]
A4["AGENTS.md<br>Unified Management Guide"]
A5["Methodology<br>Guide"]
A6["AI Credits<br>Optimization"]
Q1 -- No --> A1
Q1 -- Yes --> Q2
Q2 -- Yes --> A2
Q2 -- No --> Q3
Q3 -- Yes --> A3
Q3 -- No --> Q3b
Q3b -- Yes --> A3b
Q3b -- No --> Q4
Q4 -- Yes --> A4
Q4 -- No --> Q5
Q5 -- Yes --> A6
Q5 -- No --> A5
style A1 fill:#e1f5fe,stroke:#0288d1
style A2 fill:#e1f5fe,stroke:#0288d1
style A3 fill:#e1f5fe,stroke:#0288d1
style A4 fill:#e1f5fe,stroke:#0288d1
style A5 fill:#e1f5fe,stroke:#0288d1
style A6 fill:#fff3e0,stroke:#f57c00| Flow Result | Overview |
|---|---|
| Custom Instructions Complete Guide | Start here for first-time setup |
| applyTo Pattern Guide | Switch instructions by file path |
| Agent Skills Guide | Package procedures for autonomous execution |
| Hooks Complete Guide | Deterministic control over tool execution and audit logging |
| AGENTS.md Unified Management Guide | Manage shared rules across multiple AI tools |
| Methodology Guide | Cross-tool operational design |
| AI Credits Optimization | Keep usage-based billing under control |
Guide by Use Case¶
Getting Started & Basics¶
| What You Want to Do | Recommended Article |
|---|---|
| Teach Copilot your project rules | Custom Instructions Complete Guide |
| Apply different rules per path | applyTo Pattern Guide |
| Use common rules across multiple AI tools | AGENTS.md Unified Management Guide |
| Block dangerous operations or add audit logging | Hooks Complete Guide |
| Check personal data training settings | GitHub Copilot Data Training Policy |
Agent Utilization & Operations¶
| What You Want to Do | Recommended Article |
|---|---|
| Hand off procedures to an agent for autonomous execution | Agent Skills Guide + Agent Mode |
| Combine Copilot with Claude Code | Multi-Agent Collaboration Guide + Methodology Guide |
| Start Copilot cloud agent from internal portals or batch jobs | Copilot Cloud Agent REST API |
| Distribute Skills, Hooks, and MCP as enterprise standards | Copilot CLI Enterprise-Managed Plugins |
| Measure Copilot code review value with KPIs | Copilot Code Review Metrics |
| Understand the AI Credits migration and budget impact | GitHub Copilot AI Credits Billing |
| Optimize AI Credits usage | GitHub Copilot AI Credits Optimization |
| Check Microsoft's in-house model performance and pricing | MAI-Code-1-Flash Analysis |
| Separate billing boundaries for Skills, MCP, and external context | GitHub Copilot AI Credits Cost Design |
| Understand the downgrade criticism and lock-in risk | GitHub Copilot AI Credits Lock-in Risk |
| Optimize legacy premium request spending | Premium Request Optimization + AI Credits Optimization |
Instruction System¶
The system for conveying project knowledge to Copilot consists of four layers plus two extensions.
| Mechanism | When Applied | Typical Use |
|---|---|---|
copilot-instructions.md | Always (all requests) | Coding conventions, tech stack |
.instructions.md + applyTo | When file path matches (automatic) | Frontend/backend-specific rules |
| Agent Skills | When the agent determines it's needed | Terraform review procedures, release procedures |
| AGENTS.md | At agent startup | Project-wide policies (shared across multi-tools) |
Custom Instructions Complete Guide
How to configure
.github/copilot-instructions.mdand.instructions.md, the overall instruction system architecture, and best practicesConditional instructions based on file paths. Agent-specific control via
excludeAgentSKILL.md format, Progressive Disclosure, and on-demand procedural knowledge packages
AGENTS.md Unified Management Guide
Project instructions shared across GitHub Copilot, Claude Code, Cursor, and more, with a symlink strategy
For operational considerations, also refer to the Copilot Methodology Guide (criteria for adopting Skills / AGENTS.md, role division with Claude Code).
Agent Features¶
Copilot goes beyond code completion, functioning as an agent that autonomously executes tasks.
Multi-Agent Collaboration Guide
Overview of Agent Mode, Coding Agent, and third-party integrations, pricing structure, and latest model information
Design points for invoking cloud agent from internal portals, batch jobs, and bulk changes.
Model choice, context scope, agent work, and budget design under usage-based billing
Legacy optimization guidance for users who still have request-based billing
Measure AI review value through suggestion volume and applied rate by
securityandbug_risk.
Past update records: Agent Mode & Claude Code (August 2025 Edition)
Agent Skills also function as procedural knowledge for autonomous agent execution. For details, see the Agent Skills card in the Instruction System section.
Current billing after June 2026
The plan table below reflects the premium-request model as of February 2026. For AI Credits savings under usage-based billing after June 1, 2026, see AI Credits Optimization.
Premium request consumption varies by model. The legacy plan request counts and included models are as follows.
Pricing Plans (as of February 2026) — Free (50/month) to Enterprise (1,000/user/month). Agent Mode available on all plans
| Plan | Monthly Cost | Premium Requests | Included Models (No Request Consumption) | Primary Audience |
|---|---|---|---|---|
| Free | $0 | 50 | GPT-4o, GPT-4.1 | Individuals & Students |
| Pro | $10 | 300 | GPT-4o, GPT-4.1, GPT-5 mini | Individual Developers |
| Pro+ | $39 | 1,500 | GPT-4o, GPT-4.1, GPT-5 mini | Heavy Users |
| Business | $19/user | 300/user | GPT-4o, GPT-4.1, GPT-5 mini | Teams |
| Enterprise | $39/user | 1,000/user | GPT-4o, GPT-4.1, GPT-5 mini | Large Organizations |
- Agent Mode is available on all plans
- Premium requests are consumed when using high-performance models (Claude Opus 4.6, GPT-5.1, Gemini 3 Pro, etc.). Basic code completion is unlimited
- Additional purchase: $0.04/request
- For the latest pricing and model information, see GitHub Docs
Related Topics¶
Guides for tools and technologies commonly used alongside Copilot.
- Claude Code: Claude Code Article Guide — Autonomous agent comparison
- Codex CLI: Codex Article Guide — CLI-based agent comparison
- MCP: MCP Article Guide — External tool integration
- Skills: Skills Article Guide — Agent Skills design