An LLM wiki your coding agent writes — agent memory you can open and read. The agent does the writing; easymem does the storage, full-text search and knowledge graph. No embedding model, no vector database.
easymem turns what your coding agent reads into a local wiki of markdown pages, so the next session searches the wiki instead of reading the files again.
Search is BM25 over every page, then one hop along [[links]] — a page that
never mentions your words still comes back when a page that matches links to it.
Everything lands in plain markdown you can read, edit and commit. There is no
model inside easymem and no database: the agent writes, easymem stores, searches
and links.
npx skills add zhangxiangliang/easymem
That installs the skill into Claude Code, Cursor, Codex and other agents. From then on the AI writes down what it reads and searches the wiki first, instead of reading your files again.
No CLI? Hand your AI this line instead, and it does the rest:
Read and follow https://github.com/zhangxiangliang/easymem/blob/main/SKILL.md
No install needed. npx gets the package on first run.
npx easymem search "how does checkout work"
npx easymem --help
The index stays warm between calls. Add this and restart:
{
"mcpServers": {
"easymem": { "command": "npx", "args": ["-y", "easymem"] }
}
}
Pages land in .easymem/wiki/. Commit those; add .easymem/.state/ to
.gitignore.
lint checks, what is on disk, and what easymem does not do.The idea is Andrej Karpathy's LLM wiki: stop running RAG over raw sources on every question — have the agent compile what it reads into a wiki, then answer from the wiki. I ran that pattern in my own knowledge base for a long time. It works, it just does not work well: with nothing but a prompt, the agent writes each page in whatever shape it feels like that day, and a wiki where every page is shaped differently is a pile.
TencentDB-Agent-Memory
is the same idea with real constraints around it, and the results are a lot
better. It is also heavy — an LLM client, an HTTP API, a control panel,
multi-tenant storage, a SQLite index. easymem lifts out its MemoryKnowledge
engine — the mixed Chinese/English tokenizer, the multi-hop graph search, the
page format — and puts it behind a shell command. Nothing else.
MIT. TencentDB-Agent-Memory is MIT too.
FAQs
An LLM wiki your coding agent writes — agent memory you can open and read. The agent does the writing; easymem does the storage, full-text search and knowledge graph. No embedding model, no vector database.
The npm package easymem receives a total of 43 weekly downloads. As such, easymem popularity was classified as not popular.
We found that easymem demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.
Security News
An OpenCode user faced harassment after reporting a free AI usage loophole, drawing a bitter backlash in a GitHub thread that has since been deleted.
Threat Research
/Security News
319 browser extensions installed by 13.4 million users hijack searches, capture queries, and collect referral payments from Yahoo and Bing.
Threat Research
Malicious Firefox extensions mimic Phantom, Jupiter, and Bitget wallets to steal recovery phrases and private keys, with one fetching its payload remotely.