brainctl 2.9.1 ships: agent long-term memory in one SQLite file, stale answers 100% to 0%
R4nsack · x · 2026-10-30
brainctl 2.9.1 released: local long-term memory for AI agents in a single SQLite file—no server, no API keys, no LLM calls. Works from Python, shell, or MCP (Claude Desktop, Cursor, VS Code).
- Anti-hallucination: unanswerable searches return "not in memory" instead of the closest wrong guess
- Fact freshness: newest version of a fact supersedes old ones with history retained; stale answers dropped from 100% to 0%
- Dedup: duplicate writes merge with 0 false merges; bulk writes 5x faster
- Transparency: every number on the site links to its benchmark file
- MCP server exposes 100 tools across 27 subsystems; MIT-licensed, pip install -U brainctl
More from coding & agent
- Anthropic launches Claude Duet, a proactive agent powered by Opus 5.5 — joellewenstein · 2026-10-30
- Spring AI 2.1 adds VectorStore.upsert for apps that bring their own embeddings — therealdanvega · 2026-10-30
- Auditing an LLM-built knowledge graph: 0 of 186 facts matched verbatim, 1/3 of edges hit wrong entities — JeremyCMorgan · 2026-10-30
- Same agent harness, opposite rankings: TB4 benchmark drama exposes eval chaos — lunwang1996 · 2026-10-30
- The eternal eval debugging meme: ask the vendor if it's data or harness, it's data — dejavucoder · 2026-10-30
- DropCut: Open-Source ComfyUI Rebuild of Higgsfield Turns Photos + Song Into Beat-Synced Reels — Mediocre-Purchase-62 · 2026-10-30