BrainSync

Your AI agents forget everything between sessions. BrainSync makes them remember.

🚀 Never Miss a Fix BrainSync ships critical fixes & new features every day. Don't get left behind on a broken version. Antigravity / VS Code / Cursor / Windsurf etc: Settings → Extensions → Auto Update → ✅ Enable Antigravity IDE: Already auto-updates! ✅

💬 Join the Inner Circle Get early access to features, vote on the roadmap, and get direct support from the creator.   Members get priority bug fixes & feature requests.

🧠 Live Memory Dashboard

152K tokens optimized. 5.4 hours saved. 93 observations captured. All automatic.

📜 Recent Activity Feed

Every change — file saves, MCP calls, fixes — captured with category tags and timestamps.

🤖 Auto-Generated Skills + Custom Skills

BrainSync auto-compiles patterns into reusable skills. Create custom skills for design systems, API conventions, and more. Toggle on/off per category.

⚠️ Gotchas + Observation Detail

Color-coded by category. Full context. Never make the same mistake twice.

The Problem That's Costing You Hours Every Week

Every AI coding agent — Copilot, Cursor, Claude Code, Cline, Aider — has the same fatal flaw:

  Monday:   Agent learns your entire auth architecture     → Session ends → GONE
  Tuesday:  Same agent. Same project. Knows absolutely nothing.
  Tuesday:  You re-explain everything. Again.
  Wednesday: Repeat. Forever.

That's not pair programming. That's babysitting an amnesiac.

The average developer wastes 45 minutes per session re-explaining context to AI agents. Over a year, that's 180+ hours — gone.

BrainSync ends the loop in 30 seconds.

  Install BrainSync → Code normally → Extension captures observations silently
  → Next session, any agent in any IDE picks up exactly where you left off.

Zero config. Zero API keys. Zero cloud. Zero dollars. Just memory that persists.

The Observation Engine™ — Captures What You'd Forget

Most tools store text. BrainSync doesn't store — it observes.

The Observation Engine runs silently and captures intelligence from live sources:

Each observation is classified into 9 semantic categories — not by an LLM (which costs tokens) — but by a proprietary heuristic classifier running at $0:

The Intelligence Cortex — 20 Modules That Think, Not Just Store

This is what separates BrainSync from every "memory" tool that's really just a glorified notepad:

Universal Agent Sync — One Memory, Every IDE

BrainSync generates agent-native context files that each AI agent reads on startup — zero configuration:

Deep Integration — Any Agent, Instant Memory Access

BrainSync exposes an extensive suite of capabilities through the Model Context Protocol (MCP), including cross-project global memory, semantic code search, automated gotcha retrieval, custom skill creation, structural codebase analysis, free AI reasoning, and external documentation resolution.

Plus free AI capabilities — Offload reasoning safely with $0/query local or cloud processing.

An observation saved in Cursor is available in Claude Code 200ms later. No cloud. No webhooks. Just local file watchers.

🎯 Custom Skills — Teach Your Agents Once, They Remember Forever

BrainSync's Skill System has two modes:

Auto-Skills (fully automatic)

You code → BrainSync captures observations → 2+ similar patterns detected
  → Skill auto-compiled → Agents get rules injected automatically

Custom Skills (agent or user created)

Create professional skill packs with rules, references, and automation scripts:

.agent/skills/custom/ui-design/
├── SKILL.md          ← Rules + triggers (auto-routed by file type)
├── references/       ← Deep docs loaded on-demand
│   └── guidelines.md
└── scripts/          ← Automation scripts

Example: Tell an agent "redesign the website" → it creates a ui-design skill first with conventions ("use glassmorphism", "Inter font", "dark mode first") → then executes guided by those rules → next agent who touches CSS files automatically gets the same rules.

Skill-First Workflow

BrainSync prompts agents to plan before executing:

1. FIRST create a custom skill with conventions to follow
2. THEN execute the task guided by the skill
3. The skill auto-routes to all future agents on matching files

No more agents reinventing conventions. Define once, enforce everywhere.

Memory Vaults — Complete Project Isolation

Every project gets its own SQLite database. No cross-contamination. No shared state. No surprises.

~/.brainsync/
├── mydukan/
│   ├── memory         ← Project-specific intelligence
│   ├── vectors        ← Local semantic embeddings
│   └── skills/        ← Auto-compiled coding skills
├── brandforge/
│   ├── memory
│   ├── vectors
│   └── skills/
└── solar-app/
    ├── memory
    ├── vectors
    └── skills/

Project identity auto-detected from git remote URL or workspace folder name. Zero configuration.

Privacy — Local by Design, Not by Afterthought

We don't do cloud sync. By design.

Your code, your observations, your machine. Nothing leaves. Nothing phones home. Ever.

How BrainSync Compares

Install — 30 Seconds, Then Forget About It

1. Open VS Code, Cursor, Windsurf, or Antigravity
2. Ctrl+Shift+X → Search "BrainSync"
3. Click Install

That's it. No API key. No config file. No setup wizard.

BrainSync activates on workspace open and starts observing silently. You'll forget it's even there — until you notice your agents suddenly remember everything.

💡 Pro tip: Enable auto-update (Settings → Extensions → Auto Update) so you never miss a fix.

Commands

All optional. BrainSync works automatically. These are for when you want manual control:

Settings

Every setting is optional. Sane defaults. Zero config required.

Built By

Dawood Ahmad — building developer tools that respect your privacy and your wallet.

• 🌐 Website: brainsync.dev
• GitHub: github.com/mrd687526

License

Proprietary © 2025-2026 Dawood Ahmad

Free to use. See LICENSE.txt for full terms.