NeuraKeep
Verifiedby Agentix Labsv0.1.0
Stores source-cited agent memory with review-gated updates and failure history; exposes search/read/propose tools over local stdio MCP and a CLI, plus an OAuth-gated hosted MCP endpoint.
- Access
- MCPCLI
- Auth
- OAuth 2.0
- Skills
- 5
Skills
Memory Search
Searches stored memory for relevant entries through the memory_search tool, over stdio MCP or the local HTTP server.
Memory Read
Reads a stored memory entry in full with the memory_read tool, returning its recorded content to the agent.
Explain Memory Sources
Returns the sources behind a memory entry via memory_explain_sources, so an agent can check where a stored claim came from.
Propose Memory Update
Submits a proposed change to stored memory with memory_propose_update, queuing it for review instead of writing directly.
Apply Memory Update
Commits a reviewed memory proposal with memory_apply_update, completing the review-gated write path.
Related Agents
claude-mem
by thedotmack
Captures Claude Code tool usage as searchable observations and replays summarized context into new sessions, giving…
Graphiti
by Zep AI
Builds temporally aware knowledge graphs for agents in real time, tracking when each fact became true so memory…
Mem0
by Mem0
Persistent memory layer for AI agents and LLMs. Stores, searches, and retrieves user and session memories across…
EverOS
by EverMind
Gives agents a local-first memory layer that stores conversations, files and trajectories as Markdown, indexes them in…