The Memory Layer

Sessions end.
Knowledge remains.

Claude Code forgets. SHIVA remembers.

Cross-project. Cross-session. Cross-device.

Terminal
$ npm install -g shiva-code
$ shiva init
βœ“ 12 projects discovered
βœ“ SHIVA integrated into Claude Code
1-Minute Setup

Get Started in Three Commands

From zero to persistent memory in under a minute.

Terminal
# Installation
$ npm install -g shiva-code
# Initialize SHIVA
$ shiva init
βœ“ Config created at ~/.shiva/config.json
βœ“ 12 projects discovered
βœ“ Claude Code hook installed
# Your first memory-aware chat
$ shiva remember "Auth nutzt JWT" --category solution
βœ“ Memory saved to CLAUDE.md...
βœ“ Auto-discovers all projects
βœ“ Works with existing Claude Code
βœ“ Zero configuration needed
Session Temporary
SHIVA
Knowledge ∞ Eternal

Your sessions are temporary.
Your knowledge should be eternal.

What is SHIVA?

A persistent identity layer
for AI development workflows.

Not another tool. A memory layer that sits between you and Claude, ensuring your context, preferences, and project knowledge persist forever.

Built because re-explaining your project to AI every session is broken.

How It Works Under the Hood

Transparent architecture. No magic, just smart engineering.

Architecture
User
↓
Claude CLI / Terminal
↓
SHIVA Memory Layer
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”
↓ Context Engine
↓ Memory Store
↓ Sync Layer

CLAUDE.md Integration

Automatically writes relevant memories to CLAUDE.md. Claude reads these at every session start.

Memory Store

Local JSON files you can read, edit, and version control. Human-readable, git-friendly format.

Sync Layer (Optional)

Encrypted cloud sync for cross-device access. Opt-in only. Your data stays local by default.

What SHIVA Remembers

Six types of memory that make Claude truly understand your projects.

Rules

Project-specific guidelines and coding standards Claude should follow.

Preferences

Your personal coding style, naming conventions, and preferred patterns.

Facts

Project architecture, tech stack, file structure, and important context.

Constraints

Technical limitations, compatibility requirements, and things to avoid.

Decisions

Past architectural choices and the reasoning behind them.

Context

Current work state, recent changes, and ongoing development focus.

The Difference

See what persistent memory actually looks like.

βœ— Without SHIVA
Terminal
$ claude
> What's our database?
Claude: I don't have context about your project's database. Could you tell me what database you're using?
> What framework are we using?
Claude: I don't have that information. What framework is your project built with?
βœ“ With SHIVA
Terminal
$ claude
> What's our database?
Claude: Claude: Based on your project, you're using PostgreSQL with Drizzle ORM. Your main schema is in src/db/schema.ts.
> What framework are we using?
Claude: Claude: You're building with Astro + React, deployed on Cloudflare Workers. Your API routes are in worker/src/.

Three Powers. One Tool.

SHIVA transforms how Claude remembers.

The Third Eye

Sees what Claude forgot

Persistent memory across all sessions. Your preferences, decisions, and context - never lost again.

Many Arms

Juggles all your projects

Cross-project awareness. Claude knows what happened in Project A when working on Project B.

The Shield

Guards your data

100% local storage. No cloud dependency, no telemetry. Your knowledge stays on your machine.

The SHIVA Difference

See what persistent memory actually means for your workflow.

Feature Claude Standard Claude + SHIVA
Remembers Your Coding Preferences βœ— Every session starts fresh βœ“ Persistent across all sessions
Project Context βœ— Manual explanation required βœ“ Automatically injected
Team Knowledge Sharing βœ— Not possible βœ“ Shared memory pools
Reproducibility βœ— Varies between sessions βœ“ Deterministic responses
Audit Trail βœ— No history βœ“ Complete memory log
Cross-Project Awareness βœ— Isolated per session βœ“ Knowledge transfer enabled

SHIVA β‰  Chat History

Understanding the fundamental difference.

Chat History

  • β†’ Stores raw messages
  • β†’ Chronological order
  • β†’ Unstructured data
  • β†’ Session-bound
  • β†’ Linear retrieval
vs

SHIVA Memory

  • β†’ Extracts knowledge
  • β†’ Semantically organized
  • β†’ Typed & structured
  • β†’ Cross-session persistent
  • β†’ Contextual retrieval

Enterprise-Grade Security

Privacy-first architecture. No compromises.

End-to-End Encryption

All data encrypted at rest with AES-256. Cloud sync uses TLS 1.3 with perfect forward secrecy.

Zero-Knowledge Architecture

We can't read your memories. Encryption keys never leave your device. Not even we can decrypt your data.

Compliance Ready

SOC 2 Type II compatible. GDPR compliant. Full data export and deletion capabilities on demand.

Security Audit Schedule

Next external security audit planned for Q2 2026 with a leading security firm. All findings will be publicly disclosed.

100% Local by Default

Your memories stay on your machine. No cloud sync unless you explicitly enable it.

Zero Telemetry

No analytics. No tracking. No data collection. We don't know what you're building.

Human-Readable Files

Plain JSON you can read, edit, and version control. No proprietary formats.

Open Source

Full source code available. Audit it. Fork it. Trust but verify.

∞ Sessions remembered
0 Data sent to cloud
100% Your ownership

The Transformation

Three steps to eternal knowledge.

1

Give SHIVA Eyes

npm install -g shiva-code

Install SHIVA globally. One command. Done.

2

SHIVA Sees Your Sessions

shiva init

SHIVA scans your projects and integrates automatically into Claude Code.

3

Sessions Become Knowledge

claude

Claude never forgets again. Persistent memory activated. Cross-project awareness enabled.

Under the Hood

Technical transparency. No black boxes.

Memory Schema

Every memory is a typed, structured object with full provenance.

memory.json
"id": "mem_a7x9k2",
  "type": "preference",
  "content": "User prefers TypeScript over JavaScript",
  "confidence": 0.92,
  "source": "explicit",
  "created": "2026-01-15T10:30:00Z",
  "last_used": "2026-01-18T14:22:00Z",
  "usage_count": 47,
  "tags": ["coding", "language-preference"],
  "project_scope": "proj_frontend"
}

Conflict Resolution

When memories conflict, SHIVA applies deterministic resolution rules.

1

Explicit > Implicit

User-stated preferences always override inferred behaviors.

2

Recent > Old

More recent memories take precedence over older ones.

3

High Confidence > Low Confidence

Memories with higher confidence scores win ties.

4

Frequently Used > Rarely Used

Usage count breaks remaining ties.

Context Injection

How SHIVA enriches every prompt with relevant memory.

1

Your Prompt

"Refactor the auth module"

SHIVA Injects

Auth uses JWT Prefer async/await TypeScript strict mode
3

Enriched to Claude

"[Context: JWT auth, TypeScript, async/await] Refactor the auth module"

Debug Mode

Full transparency into what SHIVA injects.

Terminal
$ shiva search "login"
─── SHIVA Search ───
Memories found: 5
Context tokens: 847
Retrieval time: 12ms
───────────────────
β€’ [pref] Uses bcrypt for passwords
β€’ [fact] Auth endpoint: /api/auth/login
β€’ [rule] Always validate input server-side

Project Management

SHIVA manages your projects and syncs memories.

shiva sync

Store project-specific memories for consistent context.

shiva search

Search and manage memories across all projects.

.shiva/

Version-controllable config folder. Share project memories with your team via Git.

Data Flow

β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚  YOUR PROMPT                            β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                 β”‚
                 β–Ό
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚  SHIVA MEMORY LAYER                     β”‚
β”‚  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β” β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β” β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”  β”‚
β”‚  β”‚ Retrieveβ”‚ β”‚ Rank    β”‚ β”‚ Write     β”‚  β”‚
β”‚  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜ β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜ β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜  β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                 β”‚
                 β–Ό
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚  CLAUDE.md β†’ CLAUDE CODE                β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
            

Programmatic Access

Build custom integrations with the SHIVA SDK.

Coming Soon
app.ts
import { Shiva } from '@anthropic-memory/sdk';

const shiva = new Shiva({ projectId: 'proj_xyz' });

const response = await shiva.chat({
  model: 'claude-opus-4-5',
  message: 'Refactor the auth module',
  memoryScope: ['project', 'user']
});

// Access injected memories
console.log(response.injectedMemories);
// β†’ ['Uses JWT', 'TypeScript preferred', ...]
⚑ Type-safe SDK
πŸ”Œ REST + WebSocket
πŸ“¦ npm package

Team Knowledge Sharing

Shared context for collaborative development.

Team Knowledge Layer

Coding Standards Architecture Decisions API Conventions Security Policies
+

Personal Preferences Layer

Editor Settings Code Style Workflow Habits Keyboard Shortcuts

Team memories are synced via Git or SHIVA Cloud. Personal preferences stay local.

The SHIVA Principles

The philosophy behind persistent AI memory.

01

Memory is Infrastructure

Context is the foundation of intelligent AI interaction. Without memory, every conversation starts at zero.

02

Developer Control

Full transparency into what's stored, what's injected, and why. You control your AI's knowledge.

03

Privacy by Design

Encryption and local-first as the default. Your memories are yours. Period.

04

Open Standards

No vendor lock-in. Export your data anytime in standard formats. Your knowledge belongs to you.

Roadmap

The journey to universal AI memory.

Available

Claude Code Support

Full integration with Claude Code CLI. The foundation.

In Progress

Team Features

Shared memory pools for teams working on the same projects.

Planned

Multi-LLM Support

Extend persistent memory to GPT-4, Gemini, and more.

Future

Enterprise API

Full programmatic access for custom integrations.

Pricing

CLI is free. Cloud features for teams and power users.

Free

CHF 0/mo
  • CLI Tool
  • Local Memory
  • Unlimited Projects
  • Auto-Integration
Ready to get started?

Team

CHF 19/user/mo
  • Everything in Pro
  • Shared Context
  • Team Analytics
  • Admin Dashboard
Get Started

Ready to get started?

npm install -g shiva-code && shiva init