> ## Documentation Index > Fetch the complete documentation index at: https://docs.maximem.ai/llms.txt > Use this file to discover all available pages before exploring further. > ## Agent Instructions > Maximem Synap is a managed memory layer for AI agents: it ingests conversations and documents, extracts structured memory, and returns scoped context to the agent at retrieval time. > The Python and JavaScript/TypeScript SDKs are the supported way to use Synap. Write integration code with SDK calls, not raw HTTP requests. > Start with the Quickstart: https://docs.maximem.ai/getting-started/quickstart.md. Every SDK method is documented in the SDK reference: https://docs.maximem.ai/sdk-reference/overview.md. # Sound Like Me > Sound Like Me learns how you write and rewrites a rough draft in your voice, in Gmail, Slack, WhatsApp, LinkedIn, X and other sites, with one click. Sound Like Me learns how you write from messages you sent and builds a private profile called a **voice card**. Once you have one, a small Vity star appears next to the box you are typing in. Type a rough draft, click the star, and your draft is rewritten in place so it sounds like you, adjusted to the site you are on (casual in a chat, more formal in email). ## What you need * The [Vity browser extension](/vity/browser-extension.md), signed in to your Vity account. The star lives in the extension. * At least one source of your writing: **WhatsApp**, **Gmail**, your **AI chats**, or writing you paste in by hand. More sources give a richer voice, and the dashboard recommends adding WhatsApp or Gmail if you only connect AI chats, because chats show how you write to an assistant, not to people. ## Set it up **1. Open Sound Like Me** In the dashboard sidebar, expand **Apps** and click **Sound Like Me**. The first time, a short tour starts. You can replay it with **Take a tour**, and bring back the three-step **Getting started** card with **How it works**. **2. Connect your writing** Under **Writing sources**, turn on the sources you want. Click **What we use** on any row to read what it reads. * **WhatsApp**: opens **Settings**, then **Connections**, where you connect WhatsApp by scanning a QR code. It reads up to your 400 most recent sent messages, never messages you received. * **Gmail**: opens a Gmail connection window. It reads up to your 200 most recent sent emails, never your inbox. * **AI chats**: this switch is a preference, not a connection. The page says it uses your 30 most recent chats and only your own prompts. If it says "no chats in storage yet", fetch your chats first in **Data & Connectors**. Each source is read once when you generate your voice card, not continuously. You can also use **Add writing manually**: pick the app the writing is from under **This writing is from**, paste it (up to 20,000 characters at a time) or click **Upload .txt**, then click **Add to my voice**. **Clear all** removes what you added. **3. Generate your voice card** Click **Generate voice card**. It takes under a minute, and the page shows progress. If it is still running after about three minutes, you see "This is taking longer than usual". When it is done the **Voice card** panel shows a strength ring, a level (**basic**, **good** or **strong**), the number of voices and sources it was built from, and a hint for improving it. Click **Refresh** later to rebuild from your latest messages. If you switch **AI chats** on or off after building a card, a **Regenerate** button appears under **Writing sources**. **4. Tune the tone for each site** Under **Tone by channel**, drag a slider from casual to formal for **WhatsApp**, **Slack** (direct messages and channels separately), **Gmail**, **LinkedIn** and **X**. Each slider shows a label from "Very casual" to "Formal", and the page shows "Saved" when the change is stored. The sliders unlock once you have a voice card. **5. Try it in the Playground** In **Playground**, pick a scenario or choose **Write my own...**, type under **Your draft**, choose a site, and click **Rewrite in my voice** to preview how it sounds there. **6. Turn the star on in the extension** Click the Vity icon in Chrome, make sure **Sound Like Me** ("Rewrite a draft in your own voice while you type") is on, then open **Where Sound Like Me works** and switch on each site you want. Chrome asks for permission one site at a time, and the star does not run on a site you have not enabled. ## Use it 1. Click into a message box on an enabled site and type a draft of at least a few characters. 2. A star appears just outside the box. Click it (its tooltip reads "Rewrite in your voice (Sound Like Me)"). 3. While it works, your draft is covered by a shimmer and left unchanged underneath. When it finishes, the rewritten text replaces the draft. 4. Not what you wanted? Click **Undo**, or click the star again before you edit the text, to get your original back. Edit the text and click again to rewrite the new version. Sound Like Me keeps your meaning. If a rewrite would change it, you keep your draft and see "Kept your draft". ## Where it works The star can run on these sites once you enable them in **Where Sound Like Me works**: Gmail, Slack, LinkedIn, WhatsApp (web.whatsapp.com), Telegram (web.telegram.org), Messenger, Discord, X, Instagram, Facebook, Reddit, Hacker News, Threads and YouTube. The extension reads your draft only when you click the star, never while you type. If you have turned on the page summarizer's all-sites access, the picker notes that it already covers every site, and you must turn that off to choose sites one by one. ## Settings and turning it off * **Pause everywhere**: in the extension popup, switch **Sound Like Me** off. The star stays hidden until you switch it back on. * **Stop on one site**: switch that site off in **Where Sound Like Me works**. * **Stop using a source**: switch it off under **Writing sources**, or disconnect WhatsApp or Gmail in **Settings**, then **Connections**. Click **Refresh** (or **Regenerate**, if it appears) to rebuild the card without it. * **Delete your voice**: at the bottom of the Sound Like Me page, click **Delete my voice data**, type `DELETE MY VOICE` and confirm. This removes your voice card, the copy of messages kept to build it, your tone settings and any writing you added manually. It cannot be undone. Your connected sources keep their own data, so generating again rebuilds a card from them; disconnect them to stop that. > [!WARNING] > > Deleting your Vity account does not remove your Sound Like Me voice data. Use **Delete my voice data** first. To confirm that everything tied to your account is gone, email [support@maximem.ai](mailto:support@maximem.ai). ## What it uses and sends * To build your voice card: only the messages you sent from the sources you switched on, within the limits above, plus anything you added manually. * When you click the star: your draft text and the name of the site you are on. Nothing is read before you click. * Your connected inbox and chats are not read in the background after the card is built. Click **Refresh** to rebuild from newer messages. ## Troubleshooting Messages you may see next to the star: * **Build your voice card first**: you have no voice card yet. Open Sound Like Me in the dashboard and click **Generate voice card**. * **Kept your draft**: the rewrite would have changed the meaning, so your draft was left alone. Try a clearer draft. * **Busy right now**: try again in a moment. * **Couldn't reach your voice**: try again. If it keeps happening, reload the page. * **Timed out**: try again. * **Couldn't apply the rewrite**: the site's editor did not accept the text. Try again or paste your draft elsewhere. On the dashboard, **Couldn't build** with a reason means the voice card did not generate. Click **Try again**. A reason such as "no chat data found" means none of your sources had any writing to read; connect one or add writing manually. If the star never appears, check that **Sound Like Me** is on in the extension popup and that the site is enabled in **Where Sound Like Me works**. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.