Name: Jinn Code
Rating: 5 (1 reviews)
Author: Jinn Systems LLC

Real Extension · Real Screenshots

See inside the extension.

These are live screenshots from v0.1.8 running in VS Code. No mockups.

VS Code — Jinn Code panel · Ask / Plan / Agent modes

First-time setup screen

Settings — AWS Bedrock · agents · local models

Kernel Scheduler — Performance · Balanced · Custom

Interpreter DB — auto-calibrates any new model once

The Problem

Every other tool is either free-but-broken
or powerful-but-unpredictable.

Token billing shock

GitHub Copilot switched to token-based billing in June 2026. Agentic power users are reporting 10–50× cost increases. A $10/mo tool became a $750/mo surprise.

Unmanaged API bills

Cline, Continue, Kilo Code hand you an API key form on day one and leave you to manage it. A single heavy agentic session with Cline can run $200–500 in API costs. Free tool, surprise bill.

Editor migration tax

Cursor and Windsurf require you to abandon VS Code entirely. Your extensions, keybindings, and muscle memory — gone. The best AI tools shouldn't break your existing workflow.

What's Inside

Not a wrapper.
A cost-aware system.

4-agent pipeline

Thinker · Planner · Coder · Reviewer. Complexity routing means the Thinker only fires on heavy tasks — simple requests skip it entirely. No wasted tokens, no unnecessary Bedrock calls.

True VS Code extension

Lives in your sidebar. Ctrl+Shift+J opens it. No new IDE, no migration, no lost extensions. On the Marketplace right now.

Hybrid: local + AWS Bedrock

Run Ollama or LM Studio locally at zero cost. Connect your own AWS account for heavier tasks — Claude, Llama 4, DeepSeek, Nova, Mistral, Cohere. You pay Amazon's rates directly. We add nothing.

Dynamic Model Interpreter

Never seen a model before? Jinn Code probes it, learns the right format, stores the result in ~/.jinn-code/interpreters.json. Calibrates once. Reuses forever.

MCP server support

Connect any Model Context Protocol server — databases, file systems, APIs, custom tools. Your AI pipeline extends to your entire infrastructure.

Inline diff — Cursor-style

AI code appears as green/red hunks in your editor. ✔ Keep / ✖ Discard per hunk, or Keep All / Discard All. Never a surprise edit to your file.

Kernel Scheduler

Route tasks to the right model — and the right cost tier. Use local for Ask mode, Bedrock for Agent mode. Performance, Balanced, or Custom presets. Assign a different model to each agent role.

Codebase indexing

Keyword + semantic search over your entire workspace. Context injected into every agent run — active file, open files, diagnostics, git branch, project tree.

Zero telemetry. Zero middleman.

Credentials in VS Code SecretStorage, never in settings.json. Prompts go directly from your machine to AWS Bedrock or your local server — Jinn Systems is never in the path. Not now, not ever.

Transparency

Why is it free?
Here's exactly how it works.

01 · Local Models

Your hardware. Your cost: $0.

Connect Ollama or LM Studio running on your own machine. Prompts never leave your network. Jinn Code is just the interface — the compute is yours, so the cost is zero. Best for Ask mode and light tasks.

02 · AWS Bedrock

Your AWS account. Amazon's rates. Zero markup.

Connect your own AWS account. Prompts go directly from Jinn Code to AWS Bedrock — we are never in the path. You pay Amazon's published token rates. Not a penny extra comes to us.

03 · Smart Routing

Route cheap tasks local. Heavy tasks to Bedrock.

The Kernel Scheduler routes simple queries to your local model automatically. Only complex agentic runs hit AWS Bedrock. You get cloud-grade output without cloud-grade bills on every single request.

When does Jinn Code charge anything?

Only when you opt into the hosted cloud tier ($15–25/mo), which eliminates the need for an AWS account or local setup entirely. Until then, Jinn Code takes nothing. Your AWS bill and your electricity bill are the only costs — and those exist whether you use Jinn Code or not.

Join Waitlist →

The Pipeline

Four agents.
One mission.

Every request triggers a structured pipeline. A coordinated team of specialised agents, each with a defined role and its own model.

01

Thinker

Chain-of-thought reasoning. Fires on complex tasks only — no wasted tokens on simple ones.

02

Planner

Numbered file-level steps. You approve, edit, or skip before a single line is written.

03

Coder

Writes complete files. Full inline diff — keep or discard each hunk before it touches your repo.

04

Reviewer

PASS / WARN / FAIL verdict with specific, actionable fixes built into every pipeline run.

Agent colours match the real VS Code UI — ● Thinker ● Planner ● Coder ● Reviewer

ask_mode()

Single Thinker agent. Fast Q&A with full workspace context. Ideal for explanations, debugging, and quick fixes — and the cheapest mode to run. Routes to local by default.

plan_mode()

Planner produces a numbered file-level breakdown. Execute, Edit, or Discard each step. Zero code written until you approve — zero tokens spent on work you don't want.

agent_mode()

Full 4-agent pipeline. Plan → Think → Code → Review. Reserve this for heavy multi-file work — complexity routing ensures you only pay Bedrock rates when the task actually needs it.

Pricing

Free until you want
us to manage the cloud.

Local models and your own AWS Bedrock account are free today — always. The hosted subscription removes the need for any account or setup. You choose when, if ever, you upgrade.

LIVE NOW

Beta

Free

Install today. Run local models at zero cost, or connect your own AWS Bedrock account and pay Amazon's rates directly. We take nothing.

All 3 modes (Ask, Plan, Agent)
All 4 agents fully configurable
AWS Bedrock (your own account)
Local models (Ollama, LM Studio)
MCP server support
Inline diff, codebase indexing
Dynamic Model Interpreter

Install on VS Code →

Base (coming)

$15 /month

No AWS account needed. We manage the Bedrock infrastructure — flat monthly fee, no token billing surprises, just install and code.

Everything in Beta
Hosted models — no BYOK
No API keys required
Flat monthly fee
No token billing surprises

Join Waitlist

Pro (coming)

$25 /month

Everything in Base, plus priority model access, 500+ model selection, and early features. For power users who push it hard — and still pay a fixed bill.

Everything in Base
Priority model access
500+ model selection
Early feature access
Priority support

Join Waitlist

Early waitlist subscribers get first-month free when hosted plans launch.

Hosted Cloud · Coming Soon

No AWS account? No problem.

The hosted tier removes AWS entirely — flat fee, managed Bedrock, just install and code. Waitlist subscribers get first-month free. No token billing. Ever.

No spam. Launch notification + promo code only.

The Team

Built by builders.

Engineers who've shipped real products across three continents.

A product of

Jinn Systems LLC

Sharjah Media City (SHAMS) · UAE

Local models free.
AWS Bedrock at cost. Zero markup.

See inside the extension.