hindsight vs linearmemory
Agent memory that learns, not just recalls: retain/recall/reflect API over Postgres, SOTA on LongMemEval. Self-host via Docker with UI; Python/TS clients, any LLM provider. — versus — Legible persistent agent memory: records observable execution events, consolidates only validated knowledge and explains every relation — MCP server on PostgreSQL 17 with a web explorer.
Both are self-hosted agent memory over Postgres with an API. Hindsight is retain/recall/reflect, benchmarked on LongMemEval; LinearMemory is event-first and explanation-first — auditable history and explained relations over a scored recall.
| hindsight | linearmemory | |
|---|---|---|
| Stars | 22k | 1 |
| Forks | 1.7k | 0 |
| Language | Python | HTML |
| License | MIT | MIT |
| Last activity | 2 days ago | today |
| Topics | memory, agents | memory, knowledge-graphs |
| Curated connections | 9 | 3 |
hindsight — the curator's take
Pick it when you want a deployable memory *service* whose pitch is learning — agents that get better over time, not a transcript search. The LongMemEval lead was independently reproduced (Virginia Tech, Washington Post), which is more than most memory vendors offer, and the LLM side is pluggable down to Ollama/LM Studio for fully-local stacks. NOT an embedded library: you run a Docker service with Postgres and talk to it over HTTP — overkill for a single coding agent wanting session notes. The ™ and Hindsight Cloud signal a commercial trajectory; watch where the open/paid line lands.
linearmemory — the curator's take
Interesting for one design choice: memory is a chronological sequence of small observable events (reads, tool calls, decisions, errors, corrections) plus consolidated facts, decisions, procedures and outcomes — never the agent's raw reasoning transcript. Every graph edge carries direction, type, evidence, confidence and a human-readable explanation; humans browse the same store as a story, a relation map or a 3D graph. Postgres is authoritative, the graph projection is rebuildable, Docker Compose brings it up. NOT proven: one star, one author, no benchmark, no adopters yet. Watch it; don't bet on it.