Introduction
What is Synthesis and why you need persistent AI memory.
The Problem
AI agents forget everything between sessions. Every conversation starts from zero:
- You re-explain your project structure
- You remind the agent of past decisions
- You lose context on work-in-progress tasks
- Solutions to solved problems are forgotten
This isn't just inconvenient - it breaks the promise of AI assistance. An agent that can't remember your codebase is an agent that can't truly help you build.
The Solution
Synthesis gives your AI assistant persistent, structured memory through a simple 4-tool API:
- Load context at session start - know exactly where you left off
- Save progress after meaningful work - never lose work between sessions
- Capture lessons when problems are solved - build a knowledge base
- Search past solutions when stuck - recall how similar issues were resolved
How It Works
Synthesis stores your project knowledge - not your code. When you use Synthesis, it saves:
- Project context - current state, next steps, decisions
- Work logs - history of what was accomplished each session
- Lessons - bug fixes and techniques that can be reused
Your source code never leaves your machine. Only the context you explicitly create is stored in Synthesis.
Knowledge, not code. We store summaries, decisions, and lessons - never your actual source files. Your codebase stays on your machine.
The 4-Tool API
| Tool | Purpose |
|---|---|
synthesis_start | Load project context at session start |
synthesis_checkpoint | Save progress after meaningful work |
synthesis_lesson | Capture reusable knowledge |
synthesis_search | Find solutions from past lessons |
Key Benefits
- Works anywhere - Cloud-hosted API accessible from any MCP client
- Clean repos - No AI artifacts in your git history
- Cross-project - Lessons from one project help in another
- Secure - API keys are hashed, data is isolated per user