GitHub Copilot for Eclipse

GitHub Copilot for Eclipse brings AI-assisted coding to the Eclipse IDE. This is a fork maintained by SuperInstance that adds Constraint Theory integration via MCP and a multi-model BYOK routing system.

What This Fork Adds

Constraint Theory MCP Server (com.superinstance.constraint.copilot)

A bundled Eclipse plugin that registers a Python MCP (Model Context Protocol) server with Copilot. The server exposes 6 tools:

Tool	Description
constraint_snap	Snap pitch to the Eisenstein A₂ lattice
constraint_funnel	Gravitational pull toward a target constraint
constraint_diagnose	4-order Goodman diagnostic for constraint health
constraint_generate	Generate music in a given mode + terrain
constraint_render	Render notes to WAV audio
constraint_terrain_list	List available musical terrains

The MCP server is discovered automatically via ConstraintMcpProvider (implements IMcpRegistrationProvider). It locates the constraint-mcp-server Python package on PYTHONPATH and connects via stdio JSON-RPC. If the server is not found, the plugin degrades gracefully — Copilot simply won't show those tools.

Multi-Model BYOK Routing (MULTI-MODEL-DESIGN.md)

An architecture for routing code completions and chat requests to different AI models based on "constraint terrain" complexity:

• Shallow terrain (single-line completion, no context) → budget models (GLM-4-Flash, DeepSeek-Lite)
• Mid terrain (multi-line, file context) → standard models (DeepSeek-V3, GLM-4)
• Deep terrain (chat with tools, reasoning) → premium models (GPT-4o, Claude 3.5)
• Abyssal terrain (agent mode, multi-file) → flagship models (Claude 3.5 Sonnet, o3)

See MULTI-MODEL-DESIGN.md for the full design document.

Core Capabilities (from upstream)

• Code completions — Inline ghost-text suggestions as you type
• Next Edit Suggestions — Predict your next edit location and propose changes
• Agent Mode — Autonomous project-aware assistance with tool calling
• Ask Mode — Conversational AI for explanations, refactors, debugging
• MCP integration — Connect Copilot with external tools and services
• Custom Agents, Subagents, Plan Agent — Advanced agentic workflows
• BYOK (Bring Your Own Key) — Use your own API keys for Azure, OpenAI, Gemini, Groq, OpenRouter, Anthropic

Prerequisites

• Eclipse IDE
• Java 17+ (Bundle-RequiredExecutionEnvironment: JavaSE-17)
• Active GitHub Copilot subscription
• Python 3.10+ (for Constraint Theory MCP tools — optional)

Install

Eclipse Marketplace

1. Open Eclipse Marketplace → GitHub Copilot
2. Drag Install to your running Eclipse workspace
3. Restart Eclipse
4. Sign in to GitHub Copilot

Update Site

1. Help → Install New Software…
2. Add update site: https://azuredownloads-g3ahgwb5b8bkbxhd.b01.azurefd.net/github-copilot/
3. Select GitHub Copilot, complete installation, restart Eclipse

Building from Source

git clone https://github.com/SuperInstance/copilot-for-eclipse.git
cd copilot-for-eclipse
./mvnw clean verify

The build uses Maven + Tycho 4.0.13. Modules:

Module	Description
com.microsoft.copilot.eclipse.core	Core plugin: LSP client, completion, chat, BYOK, persistence
com.microsoft.copilot.eclipse.ui	UI layer: chat view, inline completion, model picker, preferences
com.microsoft.copilot.eclipse.ui.terminal	Terminal integration
com.superinstance.constraint.copilot	Constraint Theory MCP registration
com.superinstance.constraint.feature	Feature packaging for the constraint plugin
com.microsoft.copilot.eclipse.core.agent.*	Platform-specific Copilot agent binaries

Project Structure

com.microsoft.copilot.eclipse.core/src/
├── lsp/                  # LSP client + protocol types
│   ├── CopilotLanguageClient.java
│   └── protocol/         # BYOK, quota, conversation, coding-agent types
├── completion/           # Inline completion provider + suggestion management
├── chat/                 # Chat modes (Ask, Agent), custom modes, MCP config
├── nes/                  # Next Edit Suggestion provider
├── persistence/          # Conversation serialization (XML-based)
├── format/               # Language-specific formatting (Java, CDT)
└── IdeCapabilities.java  # IDE feature detection

com.superinstance.constraint.copilot/src/
└── ConstraintMcpProvider.java  # MCP server discovery + registration

Configuration

Constraint Theory MCP

Set the workspace path where the constraint ecosystem lives:

# Option 1: JVM system property
-Dconstraint.workspace=/path/to/constraint-ecosystem

# Option 2: Environment variable
export CONSTRAINT_WORKSPACE=/path/to/constraint-ecosystem

# Option 3: Default fallback
~/superinstance-workspace

The constraint-mcp-server Python package must be on PYTHONPATH within that workspace.

Multi-Model Routing

See MULTI-MODEL-DESIGN.md for the full routing configuration format (model-routing.json).

Related Repos

• constraint-theory-core — Mathematical primitives for the constraint ecosystem
• snapkit-v2 — Eisenstein lattice snap (Python)
• snapkit-js — Eisenstein lattice snap (JS/TS)
• style-dna — Musical DNA extraction and style morphing

Upstream

This is a fork of github/copilot-for-eclipse (Apache 2.0 License).

License

Apache License 2.0