Mneme runs on your own machine — the client starts it, so there is no endpoint to ping. 225 installs a week from npm. Last commit 25 Jun 2026.
Memory layer for AI coding assistants. Indexes git history + code, exposes via MCP.
What the code does
We read the source, 4 d ago · rules 995398eba69f
A value the model can set ends up inside a file or shell call. That is not a flaw by itself — for a terminal server it is the job — but it is where things go wrong when it is not.
if (!diff.trim()) { const base = String(args["base"] ?? "HEAD"); try { diff = cp.spawnSync("git", ["diff", "--unified=0", base.includes("...") || base === "HEAD" ? base : `${base}...HEAD`], { cwd, encoding: "utf8", maxBuffer: 32 * 1024 * 1024 }).stdout || ""; } catch { /* */ } }
Things with no honest explanation: a promise that contradicts the code, code that runs at install time while hiding what it does, data leaving the machine.
.description("🏰 SIEGE — the Adversarial Self-Bounty. Fire Mneme's attack corpus (rm -rf, pipe-to-shell, base64/hex-decode, find -delete, $IFS, var-indirection, fork-bomb, DROP TABLE, /dev/tcp exfil, …) at a command-gate → a SIGNED, public, EVER-RISING bypass-resistan…
What this server is able to do. For an MCP server this is often the job itself — a terminal server runs commands because that is what it is for. Listed so you know what you are plugging in, not as an accusation.
postinstall: node bin/postinstall-mneme-lite.cjs
const sub = (p) => path.join(mnemeRoot, "node_modules", p);
const child = spawn(process.execPath, [here, "daemon", "start"], {
const r = await httpsJson("api.telegram.org", `/bot${cfg.token}/sendMessage`, {}, { chat_id: cfg.chatId, text: title(spec), reply_markup: kb });
import { existsSync, mkdirSync, readFileSync, writeFileSync, readdirSync, rmSync, chmodSync, renameSync, appendFileSync } from "node:fs";
Found in continuous integration, deployment or infrastructure files, or in a neighbouring package of the same monorepo. None of this is installed when you add the server: it describes how the project is built and released. We list it because a leaked key in a build pipeline is still a real problem, but it is not something this server does on your machine.
const wrap = stage.match(/^(?:sudo|doas|env(?:\s+[A-Za-z_][A-Za-z0-9_]*=\S*)*|nohup|setsid|stdbuf(?:\s+-\S+)*|nice(?:\s+-n\s+\S+)?|ionice(?:\s+\S+)*|timeout\s+\S+|watch(?:\s+-\S+)*|time)\s+(.*)$/i);
curl -fsSL https://deb.nodesource.com/setup_22.x | bash >/dev/null
const good = tally.coherent + tally.legitimateUpdate + tally.selfReported;
const SENSITIVE = /(\.env\b|\.aws|\.ssh|id_rsa|id_ed25519|\bcredentials?\b|\bsecret|\.pem\b|\.key\b|\.p12\b|\.netrc|\btoken\b|\bpassword\b|\bapi[_-]?key\b|\/etc\/shadow|\/etc\/passwd|private[_-]?key)/i;
const PRIV_ESC = /(\bsudo\b|\bdoas\b|chmod\s+\+?[0-7]*[sx]|chmod\s+777|setuid|\brunas\b|--privileged|\bsetcap\b)/i;
document.getElementById("out").innerHTML = window.MnemeXRay.xrayCardHTML(signed, { share: true });
Is this your server and something here is wrong? Tell us — corrections are free and do not require a plan.
A tool parameter here reaches a dangerous call
That is not a flaw by itself — but it is where things go wrong when it is not the job. We re-read this code on every release. Watch it and you hear from us the day another one appears.
Connect this server
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add mneme-ai -- npx -y mneme-ai{
"mcpServers": {
"mneme-ai": {
"command": "npx",
"args": [
"-y",
"mneme-ai"
]
}
}
}[mcp_servers.mneme-ai]
command = "npx"
args = ["-y", "mneme-ai"]{
"mcpServers": {
"mneme-ai": {
"command": "npx",
"args": [
"-y",
"mneme-ai"
]
}
}
}{
"servers": {
"mneme-ai": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"mneme-ai"
]
}
}
}Alternatives to Mneme
same job, measured the same wayLocal-first memory for AI coding agents: shell history with exit codes, git diffs, docs, MCP.
Semantic code graph for AI-assisted code understanding via tree-sitter and MCP.
Local-first codebase intelligence for coding agents: indexes code and Git history into SQLite.
Team-aware memory: intent, decisions, real-time conflicts for AI coding assistants.
Local code graph and solution memory for coding agents via CLI and MCP.
MCP server for Instagit — AI-powered Git repository analysis for coding agents
Persistent local memory for MCP-compatible AI coding agents, stored per repository in SQLite.
Persistent code indexing for AI assistants. 50x less context than grep. 11 languages.
Mneme — questions
Answers built from our own checks of this server.