RRecords Labs Help Center

Browser extension

The Records Labs extension for Chrome puts your knowledge on every tab. It captures pages into your organization's review queue and lets you ask an agent about whatever you are looking at.

What it does

  • Capture a page, an article, a selection, a PDF, or a video transcript into your knowledge.

  • Capture a whole section of a documentation site, course or list, page by page, in a background tab.

  • Ask an agent a question, optionally using the current page as context, and insert the drafted answer into a form or editor on the page.

  • Act in browser: have an agent carry out a multi-step task in the tab. It pauses to ask before a step you did not clearly ask for, such as sending, buying or deleting, and it never types passwords, one-time codes or payment details.

Install

The extension is not in the Chrome Web Store yet. Your admin can give you the current build to install by hand:

  1. Unzip the extension package.

  2. Open chrome://extensions, turn on Developer mode, click Load unpacked, and pick the unzipped folder.

Until then, the Browser capture option under Knowledge → Sources offers a bookmarklet that needs no install.

Sign in with a personal API key

The extension signs in with a personal key, so everything it captures is attributed to you.

  1. In Records Labs, open Settings → Keys & Secrets. Under Personal keys, click Create key.

  2. Give it a Name and tick the scopes Retrieve, Answer and Ingest. Copy the key when it is shown; it is only shown once.

  3. In the extension, open Settings (the gear icon) → Connection, paste the key into Records Labs API key, and click Connect. You'll see "Connected to {organization}. Everything you capture saves into this organization."

  4. Make sure at least one agent has API turned on in its Channels step (an admin does this from the user menu → Agents). The extension answers through agents with API access.

The key is stored on your device only and never synced. You need the contributor role or higher to capture.

The side panel

Click the toolbar icon, or press Ctrl/⌘+Shift+K, to open the panel. It has two tabs.

Chat. Choose Answer or Act in browser, pick an agent and a model, and tick Use this page to include the current page. Set the depth, then Send. Insert into page places the answer into the field you were editing. In Act mode you get Run task, Save as task and Attach files (up to 10 MB each); the agent stops with Confirmation needed before a consequential step you did not ask for.

Capture. Pick what to save: Whole page, Article only, Selection, PDF, or Video transcript. All pages in this section walks a docs section, course or list in a background tab (up to 800 pages per run) with a Stop and save what's captured button. You can also jot a note. Set the Review state per capture; by default captures go to Review for a teammate to approve before they are used in answers.

Right-click menus offer Add this page to Records Labs, Add selection to Records Labs, and Ask Records Labs about this. Ctrl/⌘+Shift+S saves the current page.

Settings also hold Answers defaults (agent, depth, answer mode, and the category captures are saved as), Capture defaults, Act (browser tasks) options, Guardrails & trust, and an Activity list.

What is sent, and when

  • Page content is sent only when you act, and only to your organization's Records Labs API.

  • Personal details (email addresses, phone numbers, card numbers, US Social Security numbers, IBANs and secret-looking keys) are removed from captures and questions before sending. This is on by default.

  • Under Site rules, each site can be set to Allow, Ask or Block. Sites on the Hard block list are never touched, whatever other rules say; the list starts with common banking, health and sign-in sites, and you can add your own.

Companion endpoints

The extension talks to the public API on your behalf. In plain terms it uses endpoints to: ask an agent and search passages; capture a page or save a note into the review queue; read and update your capture defaults; list the agents with API access; list and run skills; pick the next step of a browser task; and confirm which organization the key belongs to. All of them live under /api/v1 and are documented in the Endpoints reference. Capture, browser tasks and skills need the browser extension plan feature, which is included in every plan.

Was this article helpful?
Related articles
API changelogDevelopersEndpoints referenceDevelopersGetting started with the Records Labs APIDevelopersMCPDevelopers