MemoryGraph provides graph-based persistent memory for AI assistants via the Model Context Protocol.
- for
- Developers building AI coding agents that need long-term, relational memory.
- pricing
- freemium
- license
- MIT
- v0.14.0MemoryGraph v0.14.0 — Correctness Releasea month ago
- v0.13.0v0.13.0 - TypeScript Port3 months ago
- v0.12.48 months ago
Key features
- Automatic memory capture — Stores solutions, patterns and decisions automatically as the agent works.
- Graph relationships — Creates typed edges (e.g., SOLVES, DEPENDS_ON) to enable multi-hop, temporal queries.
- Semantic recall — Fuzzy, relevance-based search across stored memories with recall_memories().
- Backend flexibility — Supports SQLite, FalkorDBLite, FalkorDB server, Neo4j, Memgraph and future cloud backend.
- Data portability — Migrate, export to JSON, and rollback memories between backends without lock-in.
- Cloud sync & team workspaces — Managed cloud backend shares memories across devices and teams with SSO support.
Use cases
- Recall a previously implemented rate-limiting pattern when adding a new endpoint.
- Query past timeout solutions to accelerate debugging of similar issues.
- Traverse cause-effect relationships to understand why a failure occurred.
- Share a knowledge graph of code decisions across a distributed development team.
MemoryGraph pricing
- PRO$5/month/month100,000 memories · Advanced search · Graph visualization · API access
- ULTRA$50/month/month500,000 memories · Advanced search · Graph visualization · API access
- TEAM$100/month/monthUnlimited memories · Up to 10 users · Team workspaces · SSO & SAML
MemoryGraph vs alternatives
 MemoryGraph |  Fine |  Unified MCP Server |  SuperAGI Cloud |  Pipelex | |
|---|---|---|---|---|---|
| Best for | Graph-structured memory for AI agents | General AI agent development | Broad MCP server platform | Managed AI agent execution | Repeatable AI workflow definition |
| Pricing | Freemium | Free | Subscription | Free | Free |
| DevHunt upvotes | 7 | 75 | 29 | 43 | 40 |
| Launched | Dec 2025 | Jan 2023 | Jan 2026 | Jan 2023 | Oct 2025 |
- MemoryGraph vs Fine: Focuses on building AI agents rather than providing persistent memory storage.
- MemoryGraph vs Unified MCP Server: Provides a generic MCP server with many tool integrations, not specialized graph memory.
- MemoryGraph vs SuperAGI Cloud: Offers cloud-hosted autonomous agents but does not include a graph-based memory layer.
- MemoryGraph vs Pipelex: Declarative workflow language for AI tasks; lacks built-in memory graph capabilities.
MemoryGraph FAQ
What is MemoryGraph?+
It is a persistent memory system that stores conversations, code patterns and solutions in a searchable knowledge graph for AI assistants.
How does it work?+
MemoryGraph integrates via the Model Context Protocol; agents call store_memory() and recall_memories() to write and read graph-based memories.
Is my data secure?+
All data is encrypted at rest and in transit, and the service never trains on user data. Self-hosting is also available.
Can I self-host?+
Yes, the MCP server is MIT-licensed and can be deployed with Docker, Kubernetes or manual setup.
Which AI assistants are supported?+
Any MCP-compatible assistant, including Claude (Desktop/Code) and custom GPT-based agents.
What happens when I hit a plan limit?+
The free tier prompts you to upgrade; paid plans keep data and simply require an upgrade to continue storing more memories.
Summarized by DevHunt from memorygraph.dev · Sep 27, 2026. Details may change; check the official site.
.png?width=64){kind=logo}
