GitHub Copilot is a code completion and programming AI-assistant developed by GitHub and OpenAI that provides access to AI-assisted programming tools for users of VS Code and other IDEs.
Since version 1.116, GitHub Copilot Chat is a built-in extension in VS Code. That means Copilot features such as chat, inline suggestions, and agents are now an integral part of VS Code.
🆓 Free tier
Code Completions: Up to 2,000 inline completions per month.
Chat Messages: Up to 50 chat requests per month.
AI Models: Access to select models
Codespaces: 120 core-hours
GitHub Education¶
GitHub Education is a free global program designed to help students, teachers, and schools learn and teach real-world software development.
🧑🎓 Teacher vs. Student
| Feature / Resource | Verified Teachers | Verified Students |
|---|---|---|
| Copilot Tier | GitHub Copilot Pro | GitHub Copilot Student |
| Monthly AI Credits | 1,500 credits | 200 credits |
| Credit Monetary Value | $15 USD | $2 USD |
| Inline Code Completions | Unlimited | Unlimited |
| Codespaces Compute | 180 core-hours | 180 core-hours |
| GitHub Actions | 3,000 runner minutes | 3,000 runner minutes |
🎓 How to enroll...
Add and verify your official school-issued email address in your GitHub Email Settings.
Go to the Education Benefits in Settings and click on the Start an application green button.
🛠️ Copilot in action...
Let’s get an impression about how working with a commercial full-featured AI agent feels like. For that we need a sandbox project first.
📦 PXL
| Purpose | Image processing |
| Language | C++20 |
| Type | Header-only template library |
| Repository | https:// |
🗺️ Implementation map
✨ Instruction for the AI agent
Add support for writing images in HDR format to stb.hpp using stbi_write_hdr() function that's already included in stb_image_write.h. Add unit tests for that to test_io_stb.cpp.Pros & Cons¶
Pros
✅ Convenient
✅ Fast
✅ Powerful
Cons
❌ No data privacy
❌ Network dependency
❌ Not sustainable
❌ AI fatigue
🧠 How Copilot’s memory works
Because VS Code and GitHub Copilot act as the “orchestrators” (the host infrastructure), they handle the tools that inject memory into the AI prompt before the model even processes your query. Copilot’s memory is extension-based and cross-agent, meaning that different AI models inside VS Code Copilot environment read and write to the exact same shared memory pools. The memory structure is organized by scope and context, rather than by the specific AI model you choose, in three layers:
1️⃣ User Memory
Contains your overarching personal preferences, coding styles, and general insights. Any agent model you talk to in any workspace will load the top lines of this memory pool.
2️⃣ Repository Memory
Contains rules, build commands, and architecture guidelines tailored to the workspace you currently have open. If you switch models mid-session, both models will pull from this repo-level file.
3️⃣ Session Memory
The short-term conversation context. If you switch models within the same active chat tab, the new model inherits the current transcript and session notes up to that point.
Alternatives: BYOK/BYOM¶
VS Code natively supports Bring Your Own Key / Bring Your Own Model (BYOK/BYOM). This feature allows GitHub Copilot Chat and Agent modes to route queries to remotely or locally running LLMs that are exposed via an OpenAI-compatible HTTP API. We’ll cover them next.
👻 Inline suggestions (Ghost Text)
While local models work well for Chat, Edits, and Agent Mode, GitHub Copilot’s real-time inline ghost-text autocompletion relies on custom, specialized models trained for ultra-fast response times. Local LLMs configured via (BYOK/BYOM) are generally limited to the Chat/Agent APIs rather than replacing the real-time completion engine.
If you want 100% offline inline completions + chat using local models, check the last part of this webiner: Inline Suggestions