Features Tools Permissions Install GitHub ↗
✦
Feel Free Chat Free Chat Free With All Model In the World Limitless Intelligence, Zero Cost Powering Your Code with Global AI Your Gateway to All AI Models Feel Free Chat Free
v1.1.2 · MIT License · Free Forever

The AI Agent
that lives in
VSCode

Zen turns your editor into a full AI agent. Write files, run commands, search code — powered by every LLM on Earth. No subscription. No limits.

⬇ Install Free Explore Features →
$0
Forever Free
∞
Any LLM Provider
6
Agent Tools
MIT
Open Source
scroll
DeepSeek R1 · V3
Claude 3.5 Sonnet · Haiku
Gemini 2.0 Flash · Pro
GPT-4o · o3-mini
Llama 3 · Mixtral
Qwen 2.5 · Coder
Mistral Large · Nemo
Any Provider via Elara
DeepSeek R1 · V3
Claude 3.5 Sonnet · Haiku
Gemini 2.0 Flash · Pro
GPT-4o · o3-mini
Llama 3 · Mixtral
Qwen 2.5 · Coder
Mistral Large · Nemo
Any Provider via Elara
Why Zen

Everything you need.
Nothing you don't.

Built for developers who want full control over their AI tools — without vendor lock-in or monthly bills.

⚡

Truly Free

No subscription. No usage limits. No credit card. Bring your own API keys through Elara and pay only for what you use — often nothing with free tiers.

🌐

Every LLM

Switch between DeepSeek, Claude, Gemini, OpenAI, Llama, Qwen and more in one click. Never be locked to a single provider again.

🤖

Real Agent

Zen doesn't just suggest code — it writes files, runs terminal commands, searches your codebase, and executes multi-step tasks autonomously.

🔄

Revert Anything

Every AI response creates a checkpoint. Undo any file change or command at any point in the conversation with one click.

🔒

Private by Default

Your code goes to your own Elara instance — never to a third-party cloud. Full data sovereignty, always.

📝

Full History

Search, restore, and branch from any past conversation. Date-grouped history with title, model, and workspace context preserved.

Interface

Lives in your sidebar

No new windows. No context switching. Zen is always there, right where you need it.

Visual Studio Code — zen-project
v1.1.2
📁
🔍
ZN
⧉
←
Refactor auth module
2.4k / 8k tokens
deepseek/deepseek-coder · dev@acme.com
+
U
Refactor the auth module to use JWT tokens and add refresh token support
AI
I'll refactor the auth module for you. Let me start by reading the current implementation...
READ src/auth/auth.service.ts ✓
UPDATE src/auth/auth.service.ts ✓
CREATE src/auth/refresh.service.ts ✓
EXECUTE npm run test:auth 3 passed ✓
+
⚙
🧠
Ask Zen anything...
→
Agent Tools

Six tools.
Infinite possibilities.

Every action Zen takes is shown as a live timeline in the chat — transparent, auditable, reversible.

CREATE

write_to_file

Creates a new file or completely rewrites an existing one with the AI's output

UPDATE

replace_in_file

Smart diff-based partial replacement — only changes what needs changing, preserving context

READ

read_file

Reads file content and feeds it into the AI's context for understanding and analysis

LIST

list_files

Maps out directory structure so the AI understands your project layout before acting

SEARCH

search_files

Pattern-based search across your entire workspace — finds references, usages, and definitions

EXECUTE

run_command

Runs terminal commands with live output streaming — tests, builds, installs, deploys

Dashboard

Track your
AI usage

The HomePanel gives you an at-a-glance view of your productivity, estimated savings, and which models you use most.

247
Total Chats
↑ 12 this week
2.2k
Tools Executed
↑ 94 today
$111
Estimated Savings
vs. Copilot
99.4%
API Success Rate
All systems nominal
MODEL DISTRIBUTION
DeepSeek Models50%
Claude Models30%
Gemini / OpenAI / Others20%
Control

You decide
how much AI does.

Six permission modes let you tune the agent's autonomy — from fully automated to completely manual.

bypassPermissions

Full Access

All tools execute automatically. Maximum speed for trusted workflows.

acceptEdits

Auto File Edits

File reads and writes are automatic. Commands still require your approval.

DEFAULT
auto

Auto Reads

Safe read-only tools run automatically. Writes and commands need approval.

default

Ask Every Time

Maximum control. Every single tool action requires manual confirmation.

plan

Read Only (Plan)

AI can read and plan but cannot write files or run commands.

dontAsk

Deny All

All tools are silently blocked. Pure conversation mode, no actions.

Safety Net

Undo any
AI decision.

Every AI response creates an automatic checkpoint. Disagree with what the AI did? Revert to any point in the conversation — files restored instantly.

✓ Automatic checkpoint at every response
✓ Diff view of every file change
✓ One-click restore, no manual git needed
✓ Re-send original prompt after revert
📍 CHECKPOINT #3 2 files modified
import { JwtService } from '@nestjs/jwt';
-import { AuthService } from './auth.service';
+import { AuthService, RefreshService } from './auth';
- async login(user: User) {
+ async login(user: User): Promise<TokenPair> {
+ const refresh = await this.refreshService.create(user);
+ return { access: this.jwt.sign(payload), refresh };
}
Get Started

Up and running
in 3 steps.

1
Set up Elara — the self-hosted AI proxy server that manages your API keys and routes requests to any LLM provider.
→ elara-home.vercel.app
2
Install the Zen extension in VSCode from the marketplace or build from source.
3
Open Zen from the VSCode sidebar, configure your backend URL, and start chatting with any model.
From VSIX
From Source
Prerequisites
code --install-extension zen-1.1.2.vsix
Open Source · Free Forever

Start building
with Zen today.

No credit card. No subscription. No vendor lock-in. Just you, your editor, and the AI of your choice.

⬇ Get Zen Free on GitHub Set up Elara →
⚡ VSCode Extension
🔓 MIT License
🌐 Any LLM Provider
🔒 Self-hosted
🇻🇳 Vietnamese Support