5 min lesson
Shared knowledge bases: give Max and your agents a library they can search
Learn how Frontline knowledge bases work. Source types, plan limits, daily and weekly sync, and how Max and agents query them.
A knowledge base is a searchable library of your firm’s material. You add sources. Frontline indexes them. Max and your agents query that index when someone asks a question the CRM alone cannot answer.
Harrington Family Office lives on this. The investment policy, the fee schedule, the onboarding checklist, the compliance FAQ, the public site that explains the $12M mandate process. James Harrington asks a WhatsApp agent about custody. The agent should answer from the latest policy, not from memory.
Shared knowledge bases are the ones you assign. Private ones stay with the person who created them. Shared is the model you want when more than one agent, or Max and an agent together, should read the same material.
The name and the description are not labels for you alone. They are metadata in the prompt. When Max or an agent decides which library to search, it reads the name and the description and picks the match. “Harrington IPS and fee schedule” with a description that says “investment policy, custody rules, and advisory fees” will get queried for fee questions. “Marketing site” will not. Write both as if the model is choosing a shelf.
What can I put in a knowledge base?
Sources come in a few types.
Files are the documents you already have. PDFs, Word files, text, markdown, HTML, CSV, JSON. Upload those in the product. Max can create the knowledge base and add websites or URLs. File upload still happens in the UI.
URLs are single pages you want indexed. A fee page. A specific FAQ. A published policy.
Websites are crawls. You point at a root URL or a sitemap, set how deep to go and how many pages to take, and optionally include or skip paths so the blog stays out of the compliance library.
Each source has a status you can watch: processing, completed, failed, syncing. A knowledge base with no finished sources has nothing to search yet. Add the source, wait for completed, then ask a question.
How do plans and limits work?
Your plan gates three things that matter here.
How many sources each knowledge base can hold. That cap is per knowledge base, so a large library is often several focused bases rather than one overflowing one.
Whether websites and URLs are available. Some plans support files only. On those plans you still upload documents. Adding a site or a page is a plan feature.
Whether scheduled sync is available. Some plans refresh sources on a cadence. Others leave you with a manual resync.
There is also a limit on how many shared knowledge bases the account can keep. Shared is the type you assign to Max and to agents, so that number is the one that scales the library across the team.
Max has its own plan surface as well. Shared knowledge bases for Max are a Max plan feature. If Max cannot see a shared library, check that toggle and the Max plan, then assign the base.
How does sync keep the library current?
A source is a snapshot until you refresh it.
Auto sync can run daily, weekly, or monthly, when the plan allows it. Daily fits a help center or a fee page that moves. Weekly fits a policy site. Monthly fits material that rarely changes.
Resync is the on-demand version. Pick the URL or the website and run it again. The first ingest is the initial sync. After that you get manual resyncs and the automatic ones. Website crawls keep a sync history: pages synced, failed, rejected, removed, and how long the run took.
When Harrington updates the custody section of the public site, a weekly crawl picks it up. If James is on a call tomorrow and the page already changed, resync that source now.
How do Max and agents use a knowledge base?
You assign a shared knowledge base to Max, to one or more agents, or to both. They query it. They do not dump the whole library into every turn. They read the name and description, choose the matching library, and search with a query shaped for the question.
Max has a knowledge base setting you turn on so those assigned libraries are in play. Agents get the assignment on the agent. One agent can hold several bases. A support agent might have “Policies” and “Product docs.” A qualifying agent on WhatsApp might have only “Mandate FAQ.”
For Harrington, a clean split looks like this. “Investment policy and fees” for anything a client or prospect might ask. “Internal ops” for how the team books a mandate. Assign the first to the WhatsApp qualifier and to Max. Keep the second on Max and on the internal ops agent.
Can Max create a knowledge base for me?
Yes. Tell Max what the library is for and give it the public URLs or the site to crawl. It can create the knowledge base, set the name and description, and add those web sources.
Drop the files in the product yourself. Then assign the base to Max and to the agents that should search it. Ask James’s question once in chat and once on WhatsApp. If the answer cites the wrong shelf, tighten the name and description. That is usually the fix.