doeverything

Bring your own API key from whichever model provider you already use — no subscription required. Prompts and replies travel between your browser and that provider. Compatible endpoints let you plug in self-hosted models or private deployments.

━━━━━━━━━━━━━━━━━━━━━━━━━━━ AUTOMATE ON A SCHEDULE ━━━━━━━━━━━━━━━━━━━━━━━━━━━

Set tasks to run on their own — daily, weekly, monthly, or every N minutes:

• "Every morning at 8am, check my GitHub notifications and flag what needs attention"
• "Every Monday, pull this week's analytics from the dashboard into my notes"
• "Every 30 minutes, check if the product is back in stock and ping me"

Scheduled runs happen in a background window, retry on failure, and send a desktop notification when done so you can review the results.

━━━━━━━━━━━━━━━━━━━━━━━━━━━ SKILLS & SLASH COMMANDS ━━━━━━━━━━━━━━━━━━━━━━━━━━━

Save your best prompts as reusable skills and fire them off with a slash command:

/summarize — summarize the current page in bullet points /research $company — run competitive research on a company /scrape — turn what's on screen into a structured table /download — save what you're reading to a file

Skills support argument templates ($company, $url), tool restrictions, domain-specific activation, and model overrides.

━━━━━━━━━━━━━━━━━━━━━━━━━━━ USE IT FROM OUTSIDE THE BROWSER ━━━━━━━━━━━━━━━━━━━━━━━━━━━

doeverything also runs as an MCP server. Paste the relay URL into an MCP-compatible desktop assistant, and it can browse through your Chrome — searching, reading, reporting back. Or install the local bridge, so the coding agent in your editor can read docs and click through the app you're building.

━━━━━━━━━━━━━━━━━━━━━━━━━━━ YOU STAY IN CONTROL ━━━━━━━━━━━━━━━━━━━━━━━━━━━

Four permission modes — pick whatever fits the task:

🔍 Ask every time — approve each action before it runs 📋 Follow a plan — review the agent's plan once, then let it go ✅ Allow for site — one approval per website per session ⚡ Hands-off — full automation; the agent still refuses payments, account deletion, and mass-sends unless you explicitly tell it to

It remembers what you've approved on each site, so you aren't asked twice.

━━━━━━━━━━━━━━━━━━━━━━━━━━━ PERSISTENT MEMORY ━━━━━━━━━━━━━━━━━━━━━━━━━━━

doeverything keeps what it learns. If a store's checkout needs an extra click, it writes that down and reuses the note next visit — less rediscovery, fewer wasted tokens.

• RAM memory — fast, conversation-scoped, clears on a new chat
• Persistent memory — survives restarts, stored per domain, no size limit
• Browse, export, import, and wipe memory from the settings panel

━━━━━━━━━━━━━━━━━━━━━━━━━━━ BUILT FOR REPETITIVE WORK ━━━━━━━━━━━━━━━━━━━━━━━━━━━

If you pull data from a dozen sites a week, fill the same forms all day, test user flows by hand, or copy dashboard numbers into spreadsheets, doeverything takes over the repetition — and gives you back the hours it used to eat.

━━━━━━━━━━━━━━━━━━━━━━━━━━━ PRIVACY ━━━━━━━━━━━━━━━━━━━━━━━━━━━

Your API keys are encrypted in chrome.storage.local and sent only to the model provider you pick. doeverything does not collect, store, sell, or log your conversations or browsing data — no account system, no analytics. When MCP relay mode is on, browser tool results (page snapshots, screenshots) pass through a relay server in real time and are thrown away immediately — nothing is kept or processed by doeverything's servers.

━━━━━━━━━━━━━━━━━━━━━━━━━━━

Add doeverything to Chrome. Tell it what you need. Watch it happen.