Mnemoverse
Mnemoverse provides persistent memory infrastructure for AI agents: a hosted memory API that lets an agent write a preference, decision, lesson, or context once and recall it across Claude Code, Cursor, VS Code, ChatGPT, Python, REST, and other MCP clients. Its differentiator is outcome-informed recall: feedback changes memory valence and ranking, while concept associations expand queries beyond simple vector similarity.
At a Glance
- Individual AI developers and builders
- Early-stage AI-agent startups
- Engineering and developer-tool teams
- Teams operating multi-agent systems
- +1 more
AI Tools by Mnemoverse
(1)Mnemoverse
Persistent Memory API for Agents
Discussions
No discussions yet
Be the first to start a discussion about Mnemoverse
Latest News
MCP server 0.18.0 introduced supersession support in memory_write, preserving earlier memories as history.
Mnemoverse published updated pricing with Free, Pro, Team, and Enterprise plans, including dedicated infrastructure and self-hosting by agreement.
The open-source MCP server documentation was updated to make hosted keyless installation the primary path and API-key local installation an alternative.
Enterprise self-hosting by agreement and dedicated deployment options were added to the product's deployment model.
Products & Services
Hosted API for writing, reading, and giving outcome feedback on agent memories. It uses semantic embeddings, concept associations, automatic query expansion, importance, recency, and valence-based ranking.
MIT-licensed MCP server distributed through npm and the official MCP Registry, with hosted OAuth and local API-key connection modes for Claude Code, Claude Desktop, Cursor, VS Code, Windsurf, ChatGPT, and other MCP clients.
MIT-licensed Python package on PyPI for the same write/read/feedback memory API.
One-click VS Code Marketplace extension that signs in to Mnemoverse and runs the same agent-memory service for coding workflows.
Market Position
Mnemoverse positions itself as a managed, cross-tool, MCP-first memory layer rather than a vector database or an agent runtime. It competes with Mem0, Zep, Letta/MemGPT, Cognee, Supermemory, and LangMem; its claimed distinction is that outcome feedback tunes recall through valence and learned associations, while one hosted memory works across many MCP clients. Its trade-off is that the engine is hosted and closed, whereas Mem0, Letta, Cognee, and parts of the Zep/Graphiti ecosystem offer more self-hostable/open-source paths.
Leadership
Founders
Edward Izgorodin
Founder and engineer at Mnemoverse. Public profiles describe his work across applied AI, research, principal engineering, systems architecture, TypeScript, Rust, Python, and MCP; he maintains the Mnemoverse GitHub organization and authored the company's SLoD research paper.
Executive Team
Edward Izgorodin
Founder & Engineer
Builds Mnemoverse's core API, Python SDK, open-source MCP server, integrations, benchmark work, and research; public profiles identify him as the founder and principal technical builder.
Founding Story
Mnemoverse was started to address the session-to-session amnesia of AI agents. Its initial vision was one persistent, cross-tool memory layer—write once and recall anywhere—rather than another static vector store: memories should form associations, learn from outcomes, and surface what has actually proved useful.
Business Model
Revenue Model
Usage- and subscription-based hosted API and MCP service. Customers pay for higher query quotas, atom/storage limits, rate limits, team administration, support, dedicated infrastructure, compliance features, and enterprise deployment; the MCP server and Python SDK are open source while the hosted engine is commercial.
Pricing Tiers
1,000 queries/day, 10,000 atoms, 60 requests/minute, full API access and MCP server; no credit card required.
50,000 queries/day, 500,000 atoms, 600 requests/minute, and priority support.
500,000 queries/day, 5,000,000 atoms, 3,000 requests/minute, team management, shared domains, and priority support.
Custom limits, dedicated infrastructure, SSO/SAML, audit logs, SLA, custom data residency, BAA/SOC 2 options, and self-hosting by agreement.
Target Markets
- Individual AI developers and builders
- Early-stage AI-agent startups
- Engineering and developer-tool teams
- Teams operating multi-agent systems
- Enterprise organizations requiring managed or dedicated AI-memory infrastructure
- Persistent context for coding agents across projects and sessions
- Remembering user preferences, decisions, and lessons learned
- Cross-tool memory shared across Claude Code, Cursor, VS Code, ChatGPT, and Python agents
- Multi-agent collaboration through shared memory rooms
- Production AI agents that need feedback-informed retrieval rather than static similarity search
- Enterprise agent-memory deployments requiring dedicated infrastructure and compliance controls