Covalence

Cross-Client Memory for Claude, Cursor, and MCP (macOS)

Last verified:

Visit Covalence

What is Covalence?

Covalence is a native macOS application that provides persistent, semantic, long-term memory for any AI client, including Claude, Cursor, and other MCP-compatible tools. It runs entirely locally on your Mac with your data never leaving your device, offering a private and secure solution for cross-client AI memory. The app bundles an MCP server and a local vector database into a single download, allowing any MCP client to connect to the same database with zero configuration.

Key features include Core Memories for pinning facts the AI should always know (such as your tech stack, preferences, and standing instructions), Spaces for separating memory contexts for work/personal/projects without cross-contamination, and a global hotkey capture for storing memories without switching contexts. The technical stack uses SQLite with sqlite-vec for storage, nomic-embed-text-v1.5 running on-device via CoreML for embeddings, and hybrid BM25 + vector search for retrieval. Embeddings run sub-second on Apple Silicon.

Covalence is designed for developers, AI power users, and anyone using multiple AI clients who wants to maintain persistent context across tools and sessions. It targets users who value privacy and want to avoid cloud-based solutions, Docker containers, external databases, or API keys. The app requires macOS 15+ and is approximately 261MB in size.

Covalence pricing

Pricing model: Freemium

Free to download with no limits. The app is completely free with no paid tiers, subscriptions, or usage restrictions.

Covalence pros

  • Runs entirely locally on Mac with data never leaving device
  • No API keys required
  • No network calls needed for embeddings
  • No cloud accounts or external databases required
  • Zero configuration setup
  • Cross-client memory works across Claude, Cursor, and any MCP client
  • Same database shared across all clients automatically
  • Core Memories feature pins important facts permanently
  • Spaces isolate memory contexts for different projects
  • Global hotkey capture for quick memory storage
  • Hybrid BM25 + vector search for accurate retrieval
  • Sub-second embedding speed on Apple Silicon
  • SQLite WAL mode enables concurrent client access without blocking
  • Free to download with no limits
  • Single SQLite file stores everything simply

Covalence cons

  • macOS only - no Windows or Linux support
  • Requires macOS 15 or later
  • Apple Silicon recommended for best performance
  • No cloud sync across devices
  • Approximately 261MB size mostly due to CoreML model
  • No team or collaborative features
  • No web interface or dashboard
  • New tool with unproven long-term maintenance

Frequently asked questions about Covalence

What is Covalence?

Covalence is a native macOS app that gives any AI client persistent, semantic, long-term memory. It runs locally on your Mac with your data never leaving your device, providing cross-client memory for Claude, Cursor, and MCP-compatible tools without configuration.

How does Covalence work?

Covalence bundles an MCP server and local vector database into a single download. It uses SQLite with sqlite-vec for storage, nomic-embed-text-v1.5 running on-device via CoreML for embeddings, and hybrid BM25 + vector search. Any MCP client connects to it, allowing you to store memories in one client and retrieve them from another.

Is my data private with Covalence?

Yes. Covalence runs entirely locally on your Mac - your data never leaves your device. There are no API keys, no network calls, and no cloud storage. Everything lives in a single SQLite file on your machine.

What AI clients work with Covalence?

Covalence works with any MCP-compatible client including Claude Desktop, Cursor, Windsurf, Cline, and other tools that support the Model Context Protocol. Store memory in one client and retrieve it from another using the same database.

What are Core Memories?

Core Memories are pinned facts that the AI should always know, such as your tech stack, preferences, and standing instructions. They persist across every session and every connected client, ensuring important information is never forgotten.

What are Spaces?

Spaces are separate memory contexts that let you isolate memories for work, personal use, or individual projects. This prevents cross-contamination between different contexts while using the same Covalence installation.

Does Covalence require internet?

No. Covalence runs completely offline. Embeddings run on-device via CoreML with no API keys and no network calls. All storage and retrieval happens locally on your Mac.

What are the system requirements?

Covalence requires macOS 15 or later and is approximately 261MB in size (mostly the CoreML embedding model). It works best on Apple Silicon where embeddings run sub-second, but also supports Intel Macs.

How much does Covalence cost?

Covalence is free to download with no limits. There are no paid tiers, subscriptions, or usage restrictions.

Can multiple clients use Covalence at the same time?

Yes. SQLite in WAL mode with each MCP client as a separate process allows multiple clients to read and write simultaneously without blocking. Store a memory in Claude and retrieve it from Cursor instantly.

Use cases

Browse all AI tools on NeedAnAI