Ambipath
Sign up

Personal agents are your newest channel. Turn every agent visit into a customer relationship.

Your customers now send personal agents to book, buy and reorder for them. Ambipath gives your brand its own agent to meet them, so every visit is verified, kept inside your rules, and answered with exclusive, personalized offers in your brand’s personality.

Sign up Run the free agent audit

Personal agent visits to your brand

14 a minute1,284 today
Select a visit to see inside it
Answered by your brand agent Approved by the customer in your app Stopped by your rules

Every request a personal agent makes is a visit to your brand.

Personal agents book, reorder, change and buy across every kind of brand. Each request is a chance to recognize the customer or lose them to a line of text.

  • HotelHark
    book a king room in Denver for the 14th
    Member rate$378 for two nights, 4,200 points
  • ApparelDots
    reorder my running socks, same size
    ReorderedMember free shipping, 90 points
  • AirlineSoar
    switch me to the 4pm flight
    Sent to the customer$75 fee or 7,500 miles
  • Fitness studioInstinct
    move my Thursday class to Saturday
    Moved to Saturday, 9:00Your usual bike is held
  • GroceryMuse
    add oat milk to Saturday’s delivery
    AddedMember price, $3.49
  • ElectronicsUnknown agent
    save a new card and buy the headphones
    Stopped by your rulesNew cards need the customer
  • Car rentalTab
    keep the car two more days
    ExtendedGold rate, no change fee
  • RestaurantMuse
    get my usual for 12:30
    Your usual, ready 12:38115 points to your next reward
  • BeautyDots
    restock my moisturizer
    ReorderedFree sample added for members
  • Pet suppliesInstinct
    refill the dog food subscription
    RefilledSubscriber price, $54.20

Personal agents are taking off, and the people who try them keep using them.

When a personal agent buys or books, the customer never opens your app or your site. Your brand sees only the checkout unless your own agent meets theirs.

1,700%

more daily requests from AI agents than a year ago

Daily requests from AI agents across Cloudflare’s network, September 2026. One square is last year’s volume.

73%

of new Muse users came back the next day, against 41% for ChatGPT. A week in, Muse was still ahead.

MuseChatGPT

Share of new users still active, Sensor Tower, September 2026, 81 countries. ChatGPT’s exact day 7 figure is not published.

$190B+

of US online spending through AI agents by 2030, up from $21B this year

2026: eMarketer. 2030: Morgan Stanley, 10% to 20% of US e-commerce.

Ambipath identifies each agent, checks it, briefs it, and brings your customer back to you.

  1. 1

    See which agent acted for which customer.

    Ambipath separates agent visits from customers browsing on their own. For each agent visit you see the agent and the customer who sent it.

    Most agents browse your site like any shopper and never say what they are, so your analytics count them as people today.

    Visits to your siteExample: apparel store
    Customer browsing new arrivalsCustomer
    Muse for a member, trail runners in size 9Agent
    Customer on the returns pageCustomer
    Dots for a member, reorder of running socksAgent
    Unknown agent for a member, tried to save a new cardStopped
    Agents, 4.1% of orders this monthCustomers
  2. 2

    Agents follow your rules.

    When an agent gets an order wrong, your customer blames you. You set what an agent may do alone and what goes to the customer, and Ambipath stops only the actions that break a rule.

    Every other agent keeps buying, and Studio logs each stop with the rule behind it.

    Rules for agentsYour brand
    Purchases up to $150Reorders and saved itemsAgent decidesAsk customer
    Changes with a feeFlights, stays, bookingsAgent decidesAsk customer
    Swaps when an item or seat is goneSame price or up to $10 moreAgent decidesAsk customer
    New payment methodsAdding or saving a cardAsk customerNever

    Stopped at 11:04. An agent tried to save a new card on a member account.Rule: new payment methods. The order went through on the member’s stored card.

  3. 3

    Agents find exclusive, personalized offers for the customer.

    Ambipath gives the agent this customer’s member price, points, exclusive offers and saved preferences. The agent weighs them when it books, and the customer earns points on every agent order.

    Agents see the same prices your own site shows that customer.

    What Hark can seeExample: hotel

    King room, 2 nightsDenver, the 14th to the 16th

    $458$378

    Rewards
    Earns 4,200 points. Gold member, breakfast included.
    Offer
    Late checkout at 2pm, members only.
    Preferences
    High floor, away from the elevator.
    Today
    3 king rooms left at the member rate.
  4. 4

    Agents carry your brand’s personality.

    Ambipath gives the agent your brand’s personality: how you greet a regular, what you celebrate, what you never say. The same request sounds like you and nobody else.

    MuseSame request, three brands

    reorder my running shoes, same size

    Done. Same shoes in size 9, out tomorrow. Your last pair ran 400 miles, so these have big shoes to fill.

    Your Cloudrunner 4 in size 9 is on its way, with complimentary next-day delivery for Platinum members.

    Ordered. Size 9, $89.99 with your member price. Arrives Thursday.

    RidgelineHalden & Co.BrightmartIn Muse

    Fresh legs for Saturday.

    Thank you for shopping with us again, Dana.

    You saved $20 today.

    Saturday group runFree for members, 7amJoin
    Complimentary gift wrapA Platinum perkAdd
    Running socks, 3 pack$9 for membersAdd
  5. 5

    Bring the customer to your app for the decisions that matter.

    Some choices belong to the customer. Ambipath sends them a notification that opens your own screen. They decide in your app, and the agent finishes the job.

    A change with a fee
    A flight change, a late cancel
    A choice that needs pictures
    Two rooms, two seats, two sizes
    An order over their limit
    Anything above what they let the agent spend
    A step only their phone can do
    A payment check, an ID check
    Approve your flight changeSoar found a 4:05pm aisle seat. $75 or 7,500 miles.
    9:41
    Your brand

    Change to the 4:05pm flight

    Soar found an aisle seat, Denver to Austin. Changes with a fee need you.

    Pay the change feeCard ending 4417$75
    Use milesYou have 31,2007,500
    Departs4:05pm, Denver
    Arrives7:12pm, Austin
    Seat14C, aisle
    Confirm with Face ID

Studio shows every agent visit and what it earned you.

Studio reports what the agent channel adds: visits completed, members recognized, customers coming back. Ambipath proposes rule changes from what it sees, and you approve each one.

See Studio

StudioAgent channel, last 28 daysExample: hotel, demo data
Agent visits9,8424.1% of bookings
Completed87%of agent visits
Rewards members71%of agent bookings
Agents back72%within 30 days
MuseTwo nights in Denver, member rateCompleted
DotsAdded breakfast to a stayCompleted
UnknownTried to save a new cardStopped
HarkChanged dates, $40 feeSent to customer

Proposed by AmbipathLet agents add breakfast for Gold members without asking. 412 of 430 Gold members said yes when asked.

Review

You choose what agents may see, field by field.

Set each field to share, summary or never. A summary answers the agent without the record, such as “books this hotel most months.” Agents get answers for one customer and one task, and payment details never leave your systems.

You can delete any customer’s data from Ambipath at any time. SOC 2 Type II audit targeted for early 2027.

What agents may seeYour rewards members
Points balance and statusShareSummaryNever
PreferencesShareSummaryNever
Purchase and booking historyShareSummaryNever
Member offersShareSummaryNever
Payment detailsFixedShareSummaryNever

Give personal agents a clear way to reach your assistant.

Follow the setup guide or give the generated skill to your coding agent. Ambipath checks that your published site has chat and clear connection instructions.

HoverChat on your website

Add the website snippet so visitors can ask your assistant questions using the knowledge you approve.

Publish the connection

The same snippet links to your hosted agent manifest. Ambipath also prepares entries for your website’s agent reading guides.

Check before sharing

Check your published website from Installation, then try the specific agent client your customers will use.

Connect through the guest API or a client that supports signed Personal Agent Protocol requests. See the installation guide.

See your site the way a personal agent does.

Our free audit sends a personal agent to your site on three errands: buy a product, compare two, and find a deal. In a few minutes you see whether agents get in, whether they finish each errand, and whether your site can tell them from people.

Run the free agent audit

What brands ask first

How will agents find my assistant?

Copy the website snippet from Agent → Installation and publish it on your site. It includes HoverChat and your assistant’s discovery link. Agents can ask public questions through the guest API or a signed PAP connection. Installation includes examples and a checker. Read the installation guide.

How do you know who is asking?

Website and guest API conversations are labeled Guest. A signed PAP request verifies the calling agent’s domain. Contact details shared by a visitor remain labeled as supplied; connecting an account establishes a separate verified identity.

What if an answer needs the team?

The assistant can send the question to your inbox. Your team sees the conversation and any context the visitor chose to share, then replies there. Visitors can also request follow-up and leave an email address.

Do visitors have to sign in?

No. Visitors can ask questions immediately. Sharing contact details or connecting a customer account is optional and available from the conversation menu.

What customer information will I see?

You can review questions, answers, shared context, optional contact details, and follow-up requests beside each conversation. The panel distinguishes supplied information from a connected account.

How do I start?

Sign up with your brand details and add the knowledge your assistant should use. Available free trials include $10 of inference. Email [email protected] when you need more.

Start with the task your customers send personal agents to do most.

Run the free agent audit Create a free pilot account for $10 in credits

The founders have spent more than 25 years building products that use AI and advanced machine learning to personalize engagement for hundreds of millions of customers worldwide.