FAQ & Insights

Understand the tradeoffs, performance characteristics, and common questions about Synapse.

Frequently Asked Questions

1. Is Synapse a replacement for my AI model?

No. Synapse is infrastructure. It is a specialized transmission layer that provides your existing AI model (via MCP) with the project context it needs to reason accurately.

2. Where is my data stored?

All data—code indexes, persistent memories, and knowledge graph triples—is stored in a local SQLite database on your machine. By default, this is located at ~/.synapse/storage/synapse.db.

3. Does Synapse send my code to the cloud?

No. Synapse is strictly local-first. We do not have a cloud backend, we do not collect telemetry, and we do not use external APIs for indexing or retrieval.

4. Which AI clients are supported?

Synapse works with any client that supports the Model Context Protocol (MCP). This includes:

Claude Desktop & Claude Code
Cursor & Windsurf
Cline & Continue
Gemini CLI

5. How does Synapse handle very large codebases?

Synapse uses sqlite-vec for high-performance vector search and ripgrep for ultra-fast lexical search. It uses AST-aware chunking to ensure that even in large files, only the relevant functions are surfaced to the AI.

Pros & Cons

Pros

Zero Latency: Local SQLite retrieval is significantly faster than cloud-based vector databases.
Total Privacy: Your IP, code, and memories never leave your local environment.
Context Efficiency: Tools like agent_prime and compact modes reduce token waste, giving your AI more room to "think."
Temporal Reasoning: The only system that tracks how your architecture evolved over time, not just its current state.
Offline-Ready: Full functionality without an internet connection.

Cons

Local Resources: Indexing large projects consumes local CPU and Disk space.
Model Storage: Local embedding models require ~500MB of storage.
Node.js Dependency: Requires a modern Node.js runtime (v22.6.0+) on the host machine.

Performance Tips

1. Optimize Your Index

If indexing is slow, ensure that your .gitignore is up to date. Synapse respects ignore rules and won't waste time indexing node_modules, dist, or other build artifacts.

2. Use `agent_prime`

Instead of letting your AI agent hunt for context with dozens of small tool calls, instruct it to use agent_prime once at the start of a task. This rehydrates the most important context in a single round-trip.

3. Leverage Batch Tools

For bulk operations (like importing multiple facts or relationships), use memory_store_batch or synapse_kg_manage with the ingest_markdown action. This is significantly faster than individual calls.

Next: Learn about the system's underlying Architecture.