YourMemory
agentic memory is a pruning problem, not a hoarding problem
Last verified:
What is YourMemory?
YourMemory is a biologically-inspired, local-first persistent memory engine and Model Context Protocol (MCP) server designed for AI agents. It fixes the problem where AI assistants forget everything after every session by implementing memory that works like human memory - learning, remembering, and naturally forgetting what is no longer needed using the Ebbinghaus forgetting curve.
The tool features three-layer architecture: vector search finds what you asked for, graph traversal finds what you forgot to ask for, and Ebbinghaus decay decides what survives. It offers 89.4% Recall@5 on LongMemEval benchmark, reduces token waste by 84% at 30 sessions, and supports 7 MCP clients including Claude Code, Claude Desktop, Cursor, Windsurf, Cline, Continue, and Zed. Setup requires just two commands: pip install yourmemory and yourmemory-setup.
YourMemory is for AI developers, researchers, and teams using AI agents who need persistent memory across sessions. It supports multi-agent memory with per-agent API keys, shared/private visibility controls, and optional PostgreSQL backend for teams. The visual dashboard at localhost:3033/ui shows decay bars, graph view, and per-agent tabs. It runs entirely locally with no API calls needed for embeddings (uses local spaCy model) and stores data in local DuckDB by default.
YourMemory pricing
Pricing model: Freemium
Free. The tool is open source under CC BY-NC 4.0 license, free for personal use, research, and open source projects. No API key needed. Runs entirely on your machine with local DuckDB database by default. For commercial use or team deployments, you need to reach out. Enterprise is distributed as a private build - contact required to get started. PostgreSQL support is available via pip install yourmemory[postgres] but you must host your own Postgres instance.
YourMemory pros
- 89.4% Recall@5 on LongMemEval benchmark
- Reduces token waste by 84% at 30 sessions
- Runs 100% locally - no data leaves your machine
- Only 2 commands to full setup
- Supports 7 MCP clients simultaneously
- Ebbinghaus decay with different rates by memory category
- Graph-based retrieval finds related memories query didn't match
- Multi-agent memory with shared/private visibility controls
- Per-agent API keys with ym_ prefix
- Visual dashboard at localhost:3033/ui with decay bars and graph view
- No background process or daemon to manage
- Auto-detects and configures all MCP-compatible clients
- Optional PostgreSQL + pgvector backend for teams
- Memories refresh strength when recalled - frequently used context stays alive
- Different half-lives: strategy ~38 days, fact ~24 days, failure ~11 days
- Benchmark code publicly reproducible in repo
- Free for personal use, research, and open source
- Open source under CC BY-NC 4.0 license
YourMemory cons
- Not currently available as managed cloud service
- Non-commercial license only - commercial use requires reaching out
- No flat vector store option - biologically-inspired design only
- Self-hosted only - no SaaS option
- PostgreSQL hosting must be user-provided (Fly.io, Railway, Supabase)
- Enterprise is private build only - not publicly available
- Limited to local spaCy embeddings - no custom embedding models
- CC BY-NC 4.0 restricts commercial team deployments
Frequently asked questions about YourMemory
What makes YourMemory different from other AI memory systems?
Most memory systems store flat vectors and never forget. YourMemory adds three things: Ebbinghaus decay by category where different memory types age at different rates, graph-based retrieval that surfaces related memories the query didn't directly match, and automatic pruning so stale facts never pollute your context. Built-in Claude memory is cloud-only and limited; YourMemory runs fully locally with no data leaving your machine.
Is YourMemory cloud-based or local?
YourMemory runs entirely on your local machine. Embeddings are generated with a local spaCy model with no API calls, and all storage is in a local DuckDB file by default. Nothing is sent to external servers. For teams using PostgreSQL, the database is wherever you host it - still under your control.
How does the Ebbinghaus forgetting curve work in YourMemory?
The Ebbinghaus forgetting curve models how human memory fades over time - fast at first, then slower. YourMemory applies this to each stored memory with a strength score S(t) = e^(−λt), where λ depends on the memory's category. Strategies decay slowly with ~38 day half-life, facts moderately with ~24 days, failures quickly with ~11 days. When strength drops below 0.05, the memory is pruned from retrieval. Recalling a memory refreshes its strength so frequently used context stays alive automatically.
Does YourMemory work with multiple AI clients at once?
Yes. yourmemory-setup automatically detects and configures every MCP-compatible client on your machine: Claude Code, Claude Desktop, Cursor, Windsurf, Cline, Continue, and Zed. It writes the correct MCP server config for each one. After a client restart, YourMemory is available as a tool in all of them simultaneously.
Can multiple AI agents share memory with YourMemory?
Yes. Register each agent with register_agent() to get a ym_ API key. Memories marked shared are visible to all agents; memories marked private are only visible to the storing agent. This is useful for setups where a coding agent and a review agent should share project context but keep credentials separate.
Is there a cloud-hosted version of YourMemory for teams?
Not currently - YourMemory is fully local/self-hostable by design. For team deployments, use the PostgreSQL + pgvector backend with your own Postgres instance on Fly.io, Railway, Supabase, etc. A managed cloud option may come later.
What is the license and can I use YourMemory commercially?
Non-commercial license under CC BY-NC 4.0. Free for personal use, research, and open source projects. For commercial use or team deployments, you need to reach out to the developers.
How do I set up YourMemory with my AI agent?
Two commands: first run pip install yourmemory, then run yourmemory-setup. The setup automatically configures language model, database, and MCP config for every AI client on your machine. It writes settings to ~/.claude/settings.json for Claude Code, auto-detects Claude Desktop if installed, and auto-detects Cursor, Windsurf, Cline, Continue, and Zed. Memory rules are injected into global agent context. Restart your AI client after setup and YourMemory starts automatically as an MCP server.
What benchmark results does YourMemory have?
Evaluated on three public datasets with 89.4% Recall@5 on LongMemEval, evaluated on 500 long-horizon Q&A tasks from the official LongMemEval benchmark testing storage and retrieval pipeline end-to-end. YourMemory outperforms ReadAgent at 71.2%, MemoryOS at 80.3%, and Zep at 83.1%. All benchmarks are reproducible - benchmark code is in the repo at /benchmarks. It also achieved 52% recall on the LoCoMo benchmark.
What is the visual dashboard and what does it show?
The memory dashboard is a visual UI at localhost:3033/ui that shows decay bars, graph view, and per-agent tabs. You can see memory strength over time, visualize the graph connections between memories, and switch between different agent views to see their private and shared memories.