Blog
Notes from the team building LoreKit — shared, persistent memory for AI coding agents.
- 8 min read
Your agent's memory should scale from 6 lessons to 60,000
The obvious answer to agent memory — inject the lessons — breaks the moment your store gets interesting. LoreKit spends a fixed slice of your context window on the highest-signal, de-duplicated lessons, so the read costs the same at 6 lessons or 60,000 — and always tells you what it left out. Here's how it scales, and the honest limits.
- agent-memory
- context-window
- retrieval
- 9 min read
Agent memory doesn't need embeddings — it needs a schema
Everyone reached for vector databases. But most of what a coding agent needs to remember is small, specific, and recurring — a scoped note, not a semantic blob. Here's the deterministic, no-AI data model LoreKit runs on, the rationale for every column, and an invitation to converge on a shared shape.
- agent-memory
- data-model
- deterministic
- 6 min read
Give your coding agent a memory in one command — no signup, just a folder
Your agent re-solves yesterday's problem because nothing wrote it down. The fix isn't a platform you adopt — it's one command, a folder you own, and a loop that closes on your own disk. And when a team needs the same lore, it's a free account, one install with your key, and one command to bring everything you've already learned along.
- agent-memory
- getting-started
- cli
- 9 min read
Self-healing agents are just a loop you forgot to build
Self-healing and self-improving agents sound like an ML problem. They're not. It's a plain read-fail-write loop over shared memory — no fine-tuning, no new model call. Here's how LoreKit does it, guardrails and all.
- self-healing
- self-improving
- agents