# Get started with Mneme in 60 seconds Mneme has a lot of tools. **You don't need them.** Pick the ONE line below that matches you β€” that's your first value. Everything else is there when you want it. ```bash npm i -g mneme-ai ``` --- ## πŸ§‘β€πŸ’» You write code with an AI agent (Claude Code / Cursor / Cline / Codex / Gemini …) **Approve your agent's risky actions from your phone β€” lid closed.** Your agent keeps working; the instant it hits something that needs a human *yes*, it pings your Telegram. Tap βœ…/β›” from anywhere. ```bash # create a Telegram bot once (@BotFather β†’ /newbot β†’ copy the token), then tell your AI: # "set up phone approvals, here's my token: " # your agent runs this for you β€” you never memorize a command: mneme pager autosetup --telegram-token ``` β†’ First value: close the lid, walk away, approve from your phone. [full guide](COSMIC-PAGER.md) **Or, just verify a claim right now:** ```bash mneme verify "React 19 ships server components by default" ``` ## πŸ’¬ You chat with an AI (ChatGPT / Gemini / Claude.ai / Copilot) in the browser **Truth dots beside every AI sentence** β€” green/yellow/red, so you see a hallucination as it happens. ```bash mneme polygraph autosetup --persist ``` β†’ Then install Tampermonkey once + click "Install" on the prompt. First value: open chatgpt.com β†’ dots appear. ## 🏒 You're a CEO / owner / lead β€” you want proof your AI agents are governed **A measured, signed scorecard of what every agent did** β€” harms prevented, tokens saved, all offline-verifiable (no vendor, no Mneme, needed to check it). ```bash mneme proof # the live per-agent scorecard mneme proof verify # prove the ledger wasn't edited (hash-chained) mneme axia # the signed value ledger (tokens saved, etc.) ``` ## πŸ€– You're an AI agent reading this ``` mneme boot # the taskβ†’tool decision table + what to reach for, when mneme fit # how tightly you (this agent) integrate + the native wiring mneme signal "" # the single highest-value move for the current turn ``` --- ## The one mental model Mneme is the **neutral layer** under your AI: it **verifies** (never asserts a falsehood β€” prove-or-unknown), **remembers** (across vendors), **gates** (risky actions β†’ your phone), and **proves** (every assist counted + signed). Local-first. Owned by no AI vendor. > Lost? Run `mneme atlas` (a 3 KB map) or `mneme cheatsheet`. Don't read the 988-tool list.