Multi Agent LLM

Local LLM-powered single/multi-agent chat inside VS Code — no cloud, no API key required.

Connect to LM Studio, Ollama, or any OpenAI-compatible local server and get an AI coding assistant with full file access, web search, and multi-agent collaboration.

Features

• Agent mode — AI autonomously reads files, lists directories, writes code, runs terminal commands, and searches the web via tool calls
• Single / Multi-agent toggle — switch between one LLM and Agent A/B collaboration within the same Task
• Plan / Act — Plan mode for analysis only; Act mode for file and terminal execution
• Task sessions — multiple chat tasks with sidebar CRUD, usage tracking, approval grants
• Approval system — diff review, auto-approve policies, command allowlist/blocklist, dangerous command blocking
• Checkpoints — snapshot and restore file changes per Task
• Multi-provider — OpenAI-compatible, Anthropic, Google Gemini, Ollama
• MCP & Browser — MCP server tools, Playwright browser automation
• Rules & Skills — .clinerules, /skillname on-demand skills
• Memory Bank — Cline-compatible memory-bank/*.md persistent project context
• @ Mentions — @file, @folder, @problems, @git, @editor
• Token / cost tracking — per-Task usage badge in toolbar
• Auto file apply — created files saved instantly; modified files open Diff Editor for Accept/Reject
• Git & workspace context — branch, commits, open tabs in system prompt
• Image & file attachments — vision-capable models

Supported Models (Tool Call)

Quick Start

VS Code Marketplace (권장)

1. Install from the VS Code Marketplace
2. Start LM Studio or Ollama and load a model
3. Ctrl+Shift+P → Multi Agent LLM: 채팅 패널 열기
4. Enter server URL and model name in the settings panel

VSIX 로컬 설치

1. .vsix 파일 준비 (예: vsix/multi-agent-llm-0.1.1.vsix)
2. VS Code

   code --install-extension vsix/multi-agent-llm-0.1.1.vsix
   

3. Cursor (별도 설치 필요 — code CLI는 VS Code 전용)

   cursor --install-extension vsix/multi-agent-llm-0.1.1.vsix
   

4. 필수: Ctrl+Shift+P → Developer: Reload Window (또는 VS Code/Cursor 재시작)

설치 확인 체크리스트

설치 ID: evuela.multi-agent-llm

문제 해결

Build & Package (개발자)

npm install
npm run lint          # tsc --noEmit
npm run test          # unit tests (node:test)
npm run build         # dist/extension.js
npm run package
# → vsix/multi-agent-llm-{version}.vsix

Docs: docs/getting-started.md · docs/architecture.md · docs/testing.md

dist/, node_modules/, vsix/*.vsix는 Git에 포함하지 않습니다. 클론 후 npm install && npm run compile → npm run package.

.vscodeignore로 src/, node_modules/, **/*.map 등 제외 후 dist/extension.js만 VSIX에 포함.

Settings

한국어 안내

로컬 LLM(LM Studio, Ollama 등)을 VS Code/Cursor 안에서 에이전트로 사용하는 익스텐션입니다.

주요 기능

• 단일 / 멀티 에이전트(A/B) — 같은 Task 안에서 전환
• Plan / Act, Task 사이드바, 승인·Auto-approve, Checkpoints
• Tool Call: 파일·터미널·Git·MCP·Browser·@멘션·웹 검색
• Rules / Skills / Memory Bank, 멀티 Provider, 토큰·비용 추적
• write_file 승인 UI + diff, Vision 이미지 첨부

설치 (VSIX)

# VS Code
code --install-extension vsix/multi-agent-llm-0.1.1.vsix

# Cursor
cursor --install-extension vsix/multi-agent-llm-0.1.1.vsix

설치 후 Developer: Reload Window 필수.

사용 방법

1. LM Studio 또는 Ollama 실행 후 모델 로드
2. Ctrl+Shift+P → Multi Agent LLM: 채팅 패널 열기
3. 설정(⚙)에서 URL·모델명 입력 → 연결 테스트
4. 단일/멀티 선택 후 메시지 입력

설정 키

• multiAgentLLM.singleMode.url / .model
• multiAgentLLM.agentA.url / .model
• multiAgentLLM.agentB.url / .model
• multiAgentLLM.tavilyApiKey (웹 검색, 선택)
• multiAgentLLM.maxTurns / defaultMode

패키징

npm run lint
npm run test
npm run build
npx @vscode/vsce package --no-dependencies

Publisher: evuela · 현재 버전: 0.1.1