Update the CRM from a call
The agent reads the call note in your workspace, opens the deal and fills the fields. You approve before it saves.
promptev for Chrome
promptev for Chrome is a free Chrome extension that lets an AI agent from your promptev workspace carry out tasks on websites in your own browser, asking before it acts and showing every step.
Free on the Chrome Web Store
How it works
Step 1
Install promptev for Chrome from the Chrome Web Store, then open promptev in the same Chrome once. The extension links to your own sign-in; there is nothing to configure.
One click, no keys to paste.
Then let an agent use it.
Step 2
In the promptev app, open the agent and, in Instructions & model, tick Browser under Chat capabilities. It is off by default, because an agent that clicks in your signed-in browser acts as you. A workspace admin can allow or block browser use for everyone.
Only the agents you choose can work in your browser.
Then give it a job that lives in a web page.
Step 3
In the chat, describe the task: "update the deal in the CRM from this call note". Pick how the agent checks with you: before every step, only before changes, or not at all.
You set the leash for each conversation.
Then it works in a tab of its own.
Step 4
The agent opens its own tab, reads the page and clicks or types for you. Each step shows in the side panel and in the chat. Pause it at any moment.
Done in your browser, on the record in your workspace.
What teams hand it
Much of a company's work happens in web pages that no integration reaches: portals, dashboards and admin screens. The agent works in them the way a person does, with your company's knowledge behind it.
The agent reads the call note in your workspace, opens the deal and fills the fields. You approve before it saves.
Supplier, tender or government portals with no API: the agent fills them from the contract or policy in your knowledge.
It opens the analytics or billing dashboard you are signed in to and turns the figures into a summary.
Tickets, orders or applications in a web tool: the agent checks each one against your rules and lists what needs you.
After a release, the agent clicks through the main flows in your app and reports what broke, with a picture of each step.
It reads public pages and compares them with your internal policies and price lists, which a browser alone cannot see.
Stay in control
An agent that clicks for you must be easy to watch and easy to stop. These rules are built in, not settings you have to find.
Ask every time, ask before changes (typing, or a click that sends, saves, pays or deletes), don't ask, or bypass. Chosen per conversation.
The agent works only in its own tab group, only when you gave it a task, and only while your chat is open. It does not read your other tabs.
It never types into a password field and does not read your passwords, cookies or browsing history. Browser settings pages are closed to it.
The side panel lists each step with a Pause button, and the chat keeps a picture of the page after each one.
The agent brings its knowledge, tools and approvals from your workspace. What it may read follows each person's permissions.
A workspace admin can turn browser use off for everyone. Every run is in the audit trail.
Good first tasks are the ones you know well. Keep banking and medical records for people. How the extension handles page data is in our privacy policy.
Example
After a call, a sales manager writes to the sales agent: "log this call in the CRM and set the next step to a demo on Thursday". The agent reads the call note in the workspace, opens the CRM in its own tab, finds the deal and fills the fields. The manager chose "ask before changes", so before it presses Save the chat asks for a yes. One tap, and the deal is updated, with a picture of each step in the conversation.
Questions
A Chrome extension that lets an AI agent from your promptev workspace carry out tasks on websites in your own browser, with your approval. It reads the page, clicks and types for you, and shows every step.
Yes, it is free to install. The agent's work uses your workspace's credits, like any other promptev agent. A new workspace starts with 1,000 free credits.
No. It works only in the tab the agent opens, only when an agent of your workspace was given a task, and only while your chat is open. It never types into a password field and does not read or store passwords, cookies or browsing history.
A general assistant knows the page in front of it. A promptev agent also brings your company's knowledge, tools and approval rules, answers within each person's permissions, and records every run in your workspace's audit trail. You can run it on PromptevAI or on your own model keys.
Google Chrome on Windows, macOS and Linux. Other browsers built on Chromium can often install Chrome extensions, but we test on Chrome.
Add promptev for Chrome, then give an agent its first task. New workspaces get 1,000 free credits, no card.