Everyones-AI-Lab
An AI coding agent for VS Code that works with both local and cloud LLMs. Chat, read your codebase, edit files with visual diffs, and run an autonomous multi-step agent loop — all from the sidebar.
Everyones-AI-Lab speaks a single OpenAI-compatible interface, so you can point it at a local model server (Ollama, vLLM) or a cloud provider (OpenAI, Google Gemini, Anthropic Claude) and get the same coding-companion experience.
Features
- Streaming chat — minimal first-token latency.
<think> reasoning tokens are rendered separately.
- Agent mode — read / write / edit files, run terminal commands, and drive a multi-step autonomous loop with approval UI, visual diffs, and auto-retry.
- Codebase tools —
read_file, list_files, search_files (bundled ripgrep), semantic search_codebase, LSP-powered navigation, and document export (PDF/DOCX/PPTX/XLSX).
- @mentions — pull context with
@file, @folder, @problems, @terminal, @git-changes.
- Modes — switch between Code, Architect, and Ask, each with its own tool set and rules (
.nova/rules/<mode>.md).
- Context management — adaptive compaction, message/file pinning, and file checkpoints with replay.
- Image understanding — attach images for vision-capable models.
- Korean-friendly UX — Korean-first UI, responses, and error explanations (switchable via
everyonesAiLab.language).
Privacy & data
- API keys are stored in the VS Code SecretStorage (the editor's encrypted secret store), never in plain settings.
- Audit logs (
.nova/audit/*.jsonl) and conversation logs stay on your machine — Everyones-AI-Lab does not transmit them anywhere.
- No telemetry. The extension only talks to the LLM endpoint you configure.
Supported LLM backends
| Provider |
Setting value |
Default URL |
Notes |
| Ollama |
ollama |
http://localhost:11434 |
Local model server |
| vLLM |
vllm |
http://localhost:8000 |
Local / self-hosted, OpenAI-compatible |
| OpenAI |
openai |
https://api.openai.com/v1 |
Cloud, requires API key |
| Gemini |
gemini |
Google OpenAI-compatible URL |
Cloud, requires API key |
| Anthropic |
anthropic |
https://api.anthropic.com |
Cloud (Claude), requires API key |
Installation
From the VS Code Marketplace
Search for Everyones-AI-Lab in the Extensions view (Ctrl/Cmd+Shift+X) and click Install.
From a .vsix file
code --install-extension everyones-ai-lab-<version>.vsix
Quick start
- Click the Everyones-AI-Lab icon in the Activity Bar to open the sidebar.
- Pick a provider and model in the settings (search "everyonesAiLab"), and enter an API key if your provider needs one.
- Type a message and get a streaming response.
- Use
@file / @folder to attach context, and switch modes (Code / Architect / Ask) as needed.
- When context fills up, run
/compact to summarize and free room.
Key settings
| Key |
Default |
Description |
everyonesAiLab.provider |
ollama |
ollama / vllm / openai / gemini / anthropic |
everyonesAiLab.ollamaUrl |
http://localhost:11434 |
Ollama server URL |
everyonesAiLab.vllmUrl |
http://localhost:8000 |
vLLM / custom OpenAI-compatible URL |
everyonesAiLab.temperature |
0.7 |
Sampling temperature |
everyonesAiLab.agentMode |
true |
Enable agent (file read/edit) features |
everyonesAiLab.agentAutoApprove |
false |
Auto-approve writes (caution: edits files silently) |
everyonesAiLab.language |
system |
UI language (Korean-first, English switchable) |
Search "everyonesAiLab" in the VS Code Settings UI for the full list.
Documentation
Architecture docs, PRDs, data schemas, and ADRs live under docs/ in the repository (excluded from the packaged VSIX).
docs/architecture/overview.md — architecture overview
docs/architecture/adr.md — Architecture Decision Records
docs/TROUBLESHOOTING.md — common issues and fixes
Development
pnpm install # install dependencies
pnpm run build # dual bundle build (extension + webview)
pnpm run watch # rebuild on change
pnpm test # run tests (Vitest)
pnpm run lint # lint & format check (Biome)
pnpm run package # build a distributable .vsix
License
MIT License — see the LICENSE file.
Repository
Official repository: https://github.com/terenteren/novacode — issues and pull requests are welcome.
| |