VS Code Extension

BLACKBOX AI inside your editor.

Name: BLACKBOX AI VS Code Extension
Author: BLACKBOX AI

Bring inline completions, chat-driven edits, and multi-agent execution directly into VS Code — without switching tools or breaking flow.

Install Extension Explore Extension

See it in action

What makes it powerful.

route.ts — blackbox-app

Visual Studio Code

route.ts

1import { NextRequest, NextResponse } from "next/server"

2import { auth } from "@/lib/auth"

4export async function POST(req: NextRequest) {

5 const session = await auth()

6 if (!session) return NextResponse.json({ error: "Unauthorized" }, { status: 401 })

8 const body = await req.json()

10 const result = await db.insert(tasks).values({

11 title: body.title,

12 userId: session.user.id,

13 status: "pending",

14 createdAt: new Date(),

15 }).returning()

17 return NextResponse.json(result[0])

18}

BLACKBOX AI

Copilot

Ask Blackbox...

mainLn 9, Col 1

TypeScriptUTF-8BLACKBOX AI

Feature 01

Inline completions

Get intelligent code suggestions as you type. Completions are context-aware, drawing from your open files, imports, and project structure.

Supports multi-line completions, function generation, and pattern-based suggestions. Tab to accept, Escape to dismiss — zero friction.

route.ts — VS Code

1import { auth } from "@/lib/auth"

3export async function POST(req: NextRequest) {

4 const session = await auth()

6 const result = await db.insert(tasks).values({

7 title: body.title,

8 userId: session.user.id,

9 status: "pending",

10 createdAt: new Date(),

11 }).returning()

Tab to accept

Feature 02

Chat-driven edits

Open the BLACKBOX chat panel and describe changes in natural language. Edits apply as inline diffs you can review before accepting.

Reference specific files with @path syntax. The agent understands your full workspace context for accurate, architecture-respecting changes.

api.ts — chat edit

1export async function handler(req, res) {

2 const data = req.body

3 const schema = z.object({

4 email: z.string().email(),

5 name: z.string().min(1),

6 })

7 const data = schema.parse(req.body)

9 const user = await db.users.create({ data })

10 return res.json(user)

11}

Tab to accept

Feature 03

Multi-agent execution

Run /multi-agent from the chat panel to dispatch tasks across BLACKBOX, Claude Code, Codex, and Gemini. The Chairman LLM picks the best result.

Results appear as selectable diffs in the editor. Compare implementations and choose the one that fits best — or combine elements from multiple agents.

chairman-llm — evaluate

Evaluation complete

Blackbox

9.4

Claude Code

9.1

Codex

8.7

Gemini

8.9

✓ Best result applied as inline diff

FAQ

Common Questions.

HOW DO I INSTALL IT?

Search 'BLACKBOX AI' in the VS Code Extensions Marketplace and click Install — it takes one click. The extension has over 4.2 million installations and is designed to feel like a native VS Code feature.

IS IT FREE?

The extension is free with basic features including inline completions and chat using Minimax-M2.5. Pro ($10/mo), Pro Plus ($20/mo), and Pro Max ($40/mo) unlock frontier and open-source models like Claude Opus-4.6, GPT-5.2, Llama, and Mistral, plus longer context windows.

DOES IT WORK WITH OTHER EXTENSIONS?

Yes. BLACKBOX AI integrates cleanly with existing VS Code extensions, keybindings, and themes. It uses intuitive keyboard shortcuts and a clean UI designed to minimize conflicts with your existing workflow.

WHAT LANGUAGES ARE SUPPORTED?

All languages supported by VS Code work with BLACKBOX AI. AI features are optimized for TypeScript, Python, Go, Rust, Java, C++, and all major frontend and backend frameworks.

WHICH AI MODELS ARE AVAILABLE?

The extension provides access to frontier and open-source models including Claude Opus-4.6, GPT-5.2, Gemini-3, Grok-4, Llama, and Mistral. Model availability depends on your plan tier — free users get unlimited agent requests on Minimax-M2.5.

HOW DOES MULTI-AGENT MODE WORK IN VS CODE?

Run /multi-agent from the chat panel to dispatch the same task to Blackbox, Claude Code, Codex, and Gemini simultaneously. Chairman LLM evaluates each implementation and the results appear as selectable diffs you can compare and merge.

IS MY CODE SENT TO EXTERNAL SERVERS?

Code context is sent securely over TLS 1.3 encryption for AI processing. Enterprise plans support end-to-end encryption with zero-knowledge architecture, and on-premise deployment is available for organizations requiring full data sovereignty.

Go deeper with the BLACKBOX blog.

Article

Orchestrator–Executor: A Two-Agent Split That Beats a Solo Model on SWE Tasks

Split the agent in two — a strong orchestrator that plans and verifies, a cheaper executor that implements — and Terminal-Bench 2.0 scores jump from 58.4 to 69.7.

Read article →Article

Benchmark Performance: Faster Inference, Reference-Level Model Quality

Serving a model through the BLACKBOX AI API gives higher throughput and lower latency with no observed loss in quality. Our inference stack speeds up token generation without retraining or changing model weights, and our benchmark runs show no observed regression against the published GLM 5.2, NVIDIA Ultra, and Kimi K2.7 references.

Read article →

Start with BLACKBOX

Join 30M+ developers building with BLACKBOX AI.

Get Access Read Docs