ReMe vs memoripy
Persistent memory for AI agents compared
ReMe and memoripy are Python projects that give AI agents durable local memory. ReMe centers on editable Markdown notes and flexible retrieval, while memoripy emphasizes evidence, version history, scoped records, and explanations for recalled results.
ReMe: Manage Persistent Memory for AI Agents
ReMe turns agent conversations and resources into searchable, editable Markdown memory that can evolve over time. It suits developers building assistants that need reusable knowledge across sessions and agent runtimes.
memoripy: Evidence-Tracked Memory for AI Agents
Memoripy is a local Python memory runtime for AI agents that screens incoming evidence, tracks changing facts over time, and explains recall with citations. It suits teams that need auditable, scoped memory rather than a simple vector store.
| ReMe | memoripy | |
|---|---|---|
| Language | Python | Python |
| License | Apache-2.0 | Apache-2.0 |
| Stars | 3.5k | 694 |
| Forks | 306 | 59 |
| Last analyzed | Oct 3, 2026 | Oct 9, 2026 |
Key differences
- ReMe stores memory in human-readable Markdown with frontmatter and wikilinks; memoripy preserves sourced, versioned records with temporal validity.
- ReMe captures conversations and resources, then organizes them into daily notes and longer-term entries; memoripy applies admission policies to inputs and can reject or quarantine them.
- ReMe searches passages with BM25 and wikilink expansion, with optional embeddings; memoripy retrieves across independent lanes and provides receipts explaining result selection.
- ReMe offers CLI, HTTP, MCP, and Python interfaces plus agent-runtime integrations and optional Studio; memoripy is a Python library and runtime with optional HTTP, MCP, and multi-tenant gateway tooling.
- ReMe's LLM-powered processing and optional embeddings need provider configuration; memoripy's core works without a required third-party model provider, though broader extraction can use a supplied model.
- ReMe has 3.5k stars and 306 forks; memoripy has 695 stars and 60 forks. Both use the Apache-2.0 license.
Choose ReMe if you…
- want memory stored as portable, editable Markdown files.
- need to capture conversations and resources, then consolidate them into linked notes.
- want CLI, HTTP, MCP, Python, or supported agent-runtime integrations for a shared local workspace.
Choose memoripy if you…
- need sourced records, immutable versions, and temporal validity for changing facts.
- want admission policies, memory scoping, and explanations for why records are recalled.
- need to audit evidence, conflicts, sensitive data, or retrieval feedback loops.
This comparison is generated with AI from the OSRepos analyses of both projects. Always check each project's repository and documentation before choosing.