A developer launched Fraise on Show HN, a memory database for AI agents shipped as a single binary. It stores facts as a temporal graph of facts, topics, and entities, queried with a small two-verb language (remember/recall). Recall is ranked and capped to limit the tokens the calling agent pays to read back.
Bottle is a lightweight ledger that lets AI agents log facts against custom schemas and retrieve them with a fixed set of commands. It targets the gap between unstructured markdown and full SQL, giving bots persistent, structured memory for repeated events. Developers building bots who want simple fact storage without prose or SQL complexity are the intended users.
Bottle is a lightweight ledger that lets AI agents log facts against custom schemas and retrieve them with a fixed set of commands. It targets the gap between unstructured markdown and full SQL, giving bots persistent, structured memory for repeated events. Developers building bots who want simple fact storage without prose or SQL complexity are the intended users.
Remembrane is a small library that gives LLM agents persistent memory using a single SQLite file with no dependencies. It solves the problem of storing a few thousand short strings across sessions without needing hosted APIs, vector databases, or frameworks. Its deterministic recall enables unit testing in CI.
Remembrane is a small library that gives LLM agents persistent memory using a single SQLite file with no dependencies. It solves the problem of storing a few thousand short strings across sessions without needing hosted APIs, vector databases, or frameworks. Its deterministic recall enables unit testing in CI.
Ultimate Memory is an MCP server that gives Claude Code a local memory with explicit fact validity windows, so outdated information is marked superseded rather than lingering in a vector store. It maintains a human-readable Markdown vault (Obsidian/Basic Memory) as the source of truth, solving the problem of stale facts persisting across sessions.
Ultimate Memory is an MCP server that gives Claude Code a local memory with explicit fact validity windows, so outdated information is marked superseded rather than lingering in a vector store. It maintains a human-readable Markdown vault (Obsidian/Basic Memory) as the source of truth, solving the problem of stale facts persisting across sessions.
SurrealDB founder Tobie announced the launch of SurrealDB Agent Memory, an agent memory system built on the company's multi-model database. It converts conversations and documents into structured entities, relationships, and facts with provenance and temporal history tracking, using one database for graphs, vectors, and documents instead of stitched-together storage systems. The team benchmarked on BEAM, achieving 77.2% accuracy, and is soliciting feedback from agent and context-infrastructure builders.
SurrealDB founder Tobie announced the launch of SurrealDB Agent Memory, an agent memory system built on the company's multi-model database. It converts conversations and documents into structured entities, relationships, and facts with provenance and temporal history tracking, using one database for graphs, vectors, and documents instead of stitched-together storage systems. The team benchmarked on BEAM, achieving 77.2% accuracy, and is soliciting feedback from agent and context-infrastructure builders.
Mnemo is a persistent memory layer for Claude Code that stores knowledge as a temporal knowledge graph instead of a flat vector database. It tracks when facts become true or false, enabling the AI to answer questions like 'who currently leads Solaris?' even after ownership changes. This solves the problem of outdated facts persisting in traditional vector memory systems.
Mnemo is a persistent memory layer for Claude Code that stores knowledge as a temporal knowledge graph instead of a flat vector database. It tracks when facts become true or false, enabling the AI to answer questions like 'who currently leads Solaris?' even after ownership changes. This solves the problem of outdated facts persisting in traditional vector memory systems.