Blog
Notes from the team building LoreKit — shared, persistent memory for AI coding agents.
- 7 min read
Agent memory needs its OpenTelemetry moment
Every agent tool stores memory in its own shape, and every week someone builds another one. None of them can read the others. Observability had the same problem until OpenTelemetry gave everyone one data model to share. A tour of today's memory formats, the proposals that already exist, and what a shared spec could look like.
- agent-memory
- open-format
- standards
- 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