Mnemoverse

Mnemoverse is a persistent memory API for AI agents. It is built to let a preference, decision, or lesson be written once and recalled across tools, with a single API key connecting Claude Code, Claude Desktop, Cursor, VS Code, Windsurf, ChatGPT, Python, and REST.

The service describes itself as memory for AI agents rather than a vector database. Its memory system uses Hebbian associations and Rescorla-Wagner-style outcome feedback, with HDBSCAN consolidation and Von Restorff-based forgetting. The page also describes a write-learn-recall loop: write stores a preference, decision, or lesson; learn updates valence from reported outcomes; and recall uses natural-language queries, learned associations, and valence-boosted ranking to surface what mattered before. It says an importance gate filters noise, look-alike memories can be merged, and memories that lead to good outcomes rank higher over time.

For integration, Mnemoverse offers an MCP server through the @mnemoverse/mcp-memory-server npm package. That server exposes six tools: memory_write, memory_read, memory_feedback, memory_stats, memory_delete, and memory_delete_domain. The page lists support for MCP-compatible clients including Claude Code, Claude Desktop, Cursor, VS Code + Copilot, and Windsurf, and also mentions a VS Code Marketplace listing, a PyPI package, and the official MCP Registry.

Pricing is public on the site. A free tier costs $0 and includes 1,000 queries per day, 10,000 atoms, and 60 requests per minute, with full API access and the MCP server and no credit card required. Paid plans shown are Pro at $29 per month and Team at $149 per month, and an Enterprise option is offered with custom terms, including SSO/SAML, audit logs, SLA, data residency, and BAA/SOC 2 on request. The page also says early-stage startups can start on Team free on signup, with six months if approved.