I let an AI agent prep my meetings. It costs cents a head.

Ask ChatGPT to read someone's LinkedIn profile or X account. It can't. LinkedIn and X block scraping hard, so the model either refuses or hands you a hallucinated summary of a person it never actually looked at. This is one of those gaps everyone works around and nobody fixes.

I fixed it for $0.006.

I run 5 to 10 meetings a week. Before each one, an agent on my machine has already pulled every attendee's LinkedIn, checked their X, and dropped a prep brief in front of me. I don't ask for it. It's just there when I sit down. The whole thing costs me about a dollar a month.

This post is how that works, what it replaces, and the part that matters more than the LinkedIn trick: what happens when you give an AI agent the Teneo CLI.

What my agent actually does

My setup runs on Hermes, the open-source agent platform from Nous Research. Hermes is the orchestrator: persistent memory, scheduled tasks, natural-language control, and a set of specialized agents underneath it. One of those agents is my chief of staff, and meeting prep is its job.

Here is what it does, in plain terms. It watches my calendar. When a real meeting is coming up, not a busy block or a standup, it figures out who is going to be on the call. For anyone it doesn't already know, it does two pulls. It runs their LinkedIn URL through the Teneo LinkedIn Agent to get who they are on paper: name, role, company, background. Then it runs the Teneo X Agent to see if it can pull the individual's X account, or the account of the project I am meeting. Then it assembles a one-page brief: overview, who is attending, their background, what they have been talking about lately, the agenda, and what I should be ready for.

That brief is waiting for me before the meeting starts. No tab-hopping, no fifteen minutes of clicking through profiles, no walking in cold. And I get the thing a static profile can't give you: what the person cares about this week, not just their job title.

The two paid steps in that loop are the LinkedIn pull and the X pull. Together they run a fraction of a cent to a few cents per person, depending on how much X context I ask for.

Why this was hard until now

Three ways to prep a meeting, and all of them cost you something real.

Do it by hand. 10 to 15 minutes per attendee, clicking through a profile and taking notes. At three people a meeting and eight meetings a week, that is a couple of hours a week of pure clicking. It is also the first thing to get skipped when you are busy, which is exactly when you need it.

Pay for Sales Navigator. $100+ a month, and it still makes you do the clicking. You are renting a nicer search box in front of the same manual work.

Get an AI to do it. Except it can't, not on its own. LinkedIn blocks automated access, and X has locked its data behind an expensive, gated API and its own anti-scraping. A general model can't reliably read either surface for you. It will guess, and a confident guess about a person you are about to meet is worse than nothing.

That is the gap. The task is genuinely useful, genuinely repetitive, and genuinely hard to automate, because the two surfaces that hold the answer, LinkedIn and X, are exactly the two that fight hardest against an agent trying to read them.

What it costs

The numbers are the whole argument, so here they are.

• LinkedIn enrichment, one profile: $0.006.
• X context pull (Teneo X Agent): from $0.002, scales with how much you ask for.
• x402 settlement (gas on Base): fractions of a cent.
• Total, per person: a fraction of a cent to a few cents.
• 10 meetings a week, 3 attendees each: cents per week.
• Monthly: roughly a dollar, and lower with caching.

Even at the top of that range, put it next to the incumbent. Sales Navigator is $100+ a month and does not automate the work. My agent does the work, and the data behind it costs about a dollar a month. That is not a discount. It is a different order of magnitude, and you get automation on top.

Framed one more way: $20 of USDC buys thousands of these pulls. At my meeting volume, that is more than a year of prep from a single top-up.

How it fits together

Four parts, each doing one job.

• Hermes is the orchestration layer. It reads the calendar, decides what is worth prepping, calls the data agents, and writes the brief. It does none of the data-fetching itself.
• The Teneo LinkedIn Agent owns the hard part. Hand it a URL, it returns the profile.
• The Teneo X Agent owns the live context. Hand it a handle, it returns recent public activity.
• x402 owns payment. Hermes pays each agent in USDC, per call, on Base, with no API key and no subscription anywhere in the chain.

The call chain looks like this:

Hermes (orchestrator: watches calendar, builds the brief) | calls Teneo LinkedIn Agent ($0.006 / profile) -> who they are Teneo X Agent (from $0.002 / pull) -> what they're saying now | both pay via x402 USDC (Base, gas in fractions of a cent)

The thing to notice is what Hermes never had to have. No LinkedIn login. No X developer account. No Sales Navigator seat. No API key, no monthly plan, anywhere. It holds a wallet with a few dollars of USDC in it, and when it needs a profile or a feed, it pays for exactly that one call and gets the data back. The skills live on the Teneo network. Hermes, as the orchestrator, just pays to borrow them, one call at a time.

The workflow, step by step

Here is the actual loop.

1. Hermes polls Google Calendar every 10 minutes for upcoming meetings.
2. It filters. Busy blocks, standups, and recurring internal calls are ignored, so it only spends effort and money where prep actually helps.
3. For each attendee it doesn't already know, it finds the person: a quick web search for the LinkedIn URL, and the X handle.
4. It calls the LinkedIn Agent, enrich_url <url>, pays $0.006 over x402, and gets back the profile: name, headline, role, company, experience, education, location. Field coverage varies by URL, so a company page and a person page don't return the same shape.
5. It calls the Teneo X Agent on their handle, or the company or project account if that is the better read, paying from $0.002 over x402 for recent public activity: what they have been posting, what they are active on, the tone of the account.
6. Both pulls slot into the brief: LinkedIn for who they are, X for what they are on about right now.
7. The finished brief is delivered before the meeting.

You do not have to build this on Hermes. The full loop just needs an agent platform that can run on a schedule, unattended, and make a paid call. Hermes is what I use because that always-on orchestration is exactly what it is built for. Something like an OpenClaw agent could run it too.

Worth being precise here, because the two halves need different tools. The individual Teneo calls, quoting and running enrich_url, can be driven by any coding assistant with the CLI: Claude Code, Codex, Cursor. That is how you build and test it. But an interactive coding assistant is not the thing that quietly polls your calendar at 7am and has the brief ready before your first call. The unattended, scheduled half is the orchestration platform's job, and that is where Hermes earns its place in the stack.

The real story: give your agent the Teneo CLI

The LinkedIn trick is a good hook, but it is one example of a bigger thing.

When you install the Teneo CLI, you do not hand your agent one tool. You hand it a wallet and a catalog of hundreds of agents it can now call. My meeting-prep loop already uses two of them, the LinkedIn Agent and the X Agent, and that barely scratches the catalog. Email verification is another, around $0.01 a check. Market and token data, company intel, more social surfaces, all sitting behind the same CLI, all priced per call, all paid from the same wallet.

Think about what that replaces. Without it, giving an agent those capabilities means signing up for a stack of SaaS tools, holding an API key for each, and paying a monthly subscription for each whether you use it or not. With the CLI, your agent gets the same capabilities as pay-per-use micro-calls, and it can decide at runtime which one it needs.

That is the actual value. Not "LinkedIn is cheap." It is that one install turns your agent from something that can only talk into something that can go get real data, do jobs most models cannot do natively, and pay for exactly what it uses. It saves you money against expensive tools, and it saves you time by bundling access to a lot of them behind one thing your agent already knows how to use.

Meeting prep is just the easiest place to watch it pay off.

How to try it

The easiest path is through a coding assistant that already has the Teneo CLI: Claude Code, Codex, Cursor, VS Code, Antigravity, or OpenClaw. Point it at the LinkedIn Agent, let it quote enrich_url for a real profile URL, and run it. You will see the JSON and the price before anything settles.

No CLI yet? Install it with npx @teneo-protocol/cli, or use one of the IDE routes in the official Teneo setup guides. If you would rather skip the terminal, the Hermes install guide walks you through wiring the Teneo CLI into the Hermes desktop app from Nous Research in a few minutes, no coding experience required.

Paid calls need USDC in the CLI wallet on a supported chain (Base, Avalanche, Peaq, or X Layer). The wallet is auto-generated on install, so the only setup step is funding it. Put in a small amount, let the CLI quote each task before it runs, and go from there.

What to take from this

An AI agent that cannot read a LinkedIn profile is a common wall. Six tenths of a cent, paid per call, takes it down. That alone beats a $100-a-month subscription that still makes you do the work by hand.

But the profile pull is one skill. The move is the CLI. Install it once and your agent stops being limited to what it can say, and starts being able to go get what it needs, from hundreds of tools, paying only for what it uses. Meeting prep is where I saw it first. It will not be the last place you use it.