⚡ Now in alpha — Claude Code + OpenCode supported

Your AI remembers
where you left off.

SessionGraph compresses your AI coding sessions by up to 84% and saves context across sessions — so you never re-explain your codebase again. Works with Claude Code and OpenCode out of the box.

Free to use · No account required · Runs entirely on your machine

SessionGraph is running
Live
Tokens saved today0
Cost saved today$0.3100
Sessions restored3
Compression ratio84%

Current session: sessiongraph-core

Working on /v1/responses endpoint, Axum streaming — 2,400 tokens injected

Features

Everything you need,
nothing you don't.

84%

84% compression

Context compressed before every API call. Pay for signal, not noise.

🧠

Session memory

Project state saved as a structured graph. Restored automatically next session.

📊

Live savings dashboard

See exactly what you've saved. Tokens, dollars, sessions. Updated in real time.

Zero configuration

Two env vars. Works with Claude Code, OpenCode, Aider, Continue, Codex CLI.

🔒

Fully local

Your API keys and code never leave your machine. Privacy by architecture.

🔌

Works everywhere

Any AI tool that speaks HTTP_PROXY: Claude Code, Cursor, Aider, Codex CLI, Continue, OpenCode.

84%
avg compression on Claude Code sessions
$47
saved in first month (median Pro user)
12
sessions restored per week on average

Based on internal testing across 30 coding sessions. Your results will vary by session length and tool usage pattern.

How it works

Three steps. Zero friction.

01

Install & forget

Set two environment variables. Every AI tool on your machine routes through SessionGraph automatically. Open a new terminal and it just works.

export ANTHROPIC_BASE_URL=http://localhost:4200
export OPENAI_BASE_URL=http://localhost:4200/v1

# Or run the installer — it sets these for you.
02

SessionGraph works silently

Compresses context before it hits the API. Extracts session state when you stop working. Injects it when you start again. Latency overhead: under 1ms.

# What SessionGraph does per request:
# 1. Receive request from your AI tool
# 2. Compress repeated context (avg 84%)
# 3. Forward compressed request upstream
# 4. Stream response back byte-for-byte
# 5. Parse token counts from SSE stream
# 6. Update savings dashboard in real time
03

Your AI picks up where it left off

Open Claude Code on a new day. It already knows your architecture, your decisions, your open issues. No re-explaining. The session graph is injected automatically.

# Session graph injected on next start:
{
  "where_you_left_off": "Working on Axum...",
  "decisions": [
    "Use SSE stream tee for token counting",
    "WAL mode for SQLite concurrency"
  ],
  "active_files": ["proxy/server.rs", ...],
  "conventions": ["error via anyhow::Result", ...]
}

Pricing

Simple, transparent pricing.

Free
Try it risk-free. No account required.
$0forever
Download Free
  • 3 session saves per month
  • Token compression (unlimited)
  • Live savings dashboard
  • Works with Claude Code & OpenCode
Most Popular
Pro
For developers who code with AI every day.
$15/month
Start Pro
  • Unlimited session saves
  • Token compression (unlimited)
  • Cross-device sync
  • Session search
  • Graph editing
  • Priority support
Coming Q3 2026
Team
Share context across your whole team.
$49/seat/month
Join waitlist
  • Everything in Pro
  • Shared team context
  • Centralised billing
  • Per-seat pricing
  • SSO / SAML (planned)
  • Priority support + SLA

FAQ

Common questions.