--- title: "Pine AI teardown: what runs when it calls a company" description: "Pine AI places calls from its servers, hands browser work to your Mac, and exposes 26 tool names with no schemas. What that tells you about it." url: https://personalagentbench.com/agents/pine/ last_updated: 2026-10-26 --- # Pine AI teardown: what runs when it calls a company Phones companies on your behalf about bills, refunds and cancellations. Product **19pine.ai** · Last verified **2026-10-26** · September 2026 own-draw result **20% of tasks, 95% CI 14 to 25, n=167; 2026-09-22, adapted runtime with substituted exam model**. [Provenance: differing scored subsets; first-batch records unpublished](https://personalagentbench.com/results/). Pine AI is a personal agent best known for one promise: hand it a bill dispute, and it picks up the phone so you don't have to. The marketing is about the outcome. This page is about the machine underneath, because the machine explains both how Pine approaches bill disputes and why it shows you so little about how it did it. - The model, the instructions, the tool contracts and the calls all live on Pine's servers. The one exception is browser work on the Mac app, which runs on your Mac. - The client sees 26 tool names and not one schema. One of those tools sends a fax, which tells you more about the target customer than the homepage does. - Calls made from the app keep no verbatim transcript. You get the merchant, the number, the duration, the credit cost and a summary. Calls placed through Pine's developer API do return a turn-by-turn transcript. - Memory is a server-side fact layer, not a file. We found no directly editable local memory file; corrections must go through the product. - One brand covers four different request shapes: a task assistant, a voice copilot, a REST API other people's agents can call, and a leased cloud desktop. The basics, from Pine's own pages. Its [terms of service](https://www.19pine.ai/terms-of-service) (dated 31 August 2026) name the operator as 19Pine Pte. Ltd. and put the agreement under Singapore law; on Apple's [App Store listing](https://apps.apple.com/us/app/pine-ai-assistant-agent/id6746403769) the seller is 19Pine Corp. You get it as an iPhone app, on [Google Play](https://play.google.com/store/apps/details?id=com.pineai.app.release), on the web at [19pine.ai](https://www.19pine.ai/) and on the Mac. Bills are no longer the whole pitch: Pine also sells [personal calls and messages](https://www.19pine.ai/send-a-message), where it rings a friend or relative with a message from you, and [Pine Work](https://www.19pine.ai/work), which it describes as giving your agent a computer of its own, with downloads for macOS and Windows. We examined the bill-dispute product, not those two. We took the client apart. Everything below is either something the product states outright or something we reconstructed from several pieces of evidence that agree. Where we couldn't get to the bottom of something, we say so instead of guessing. The US [App Store listing](https://apps.apple.com/us/app/pine-ai-assistant-agent/id6746403769), checked 12 October 2026, showed 4.5 out of 5 from 147 ratings. Those are user ratings, not a measured completion rate or our assessment of Pine’s reliability. ## How Pine AI works: the model and the calls run on Pine's servers Open the web app or the Mac app and you're looking at a display, not a brain. The model never runs locally, and the calls are placed from Pine's side; the exception, browser work on the Mac app, gets its own section below. A server builds the whole request, meaning instructions, retrieved facts, tool contracts, conversation history and your latest message, and the client only ever receives the results over a REST endpoint and a socket connection. That sounds like a boring implementation detail. It isn't, and here's why: you cannot see what Pine was told. The instructions that shape its behaviour never reach your machine, so there's no version of "check what it was asked to do" available to you. We probed for the production instruction text roughly 126 times, statically and at runtime. Empty every time. From outside we cannot tell whether that is a deliberate choice, but the effect is the same either way, and it's worth knowing about a product you're handing your account details to. One small thing that trips people up: the task cards on the Pine homepage each carry a field called `prompt`, and it's tempting to read that as the system instruction. It's your own message. The practical consequence is that every question about Pine's behaviour has to be answered from the outside. You can watch what it does. You cannot read what it was told to do. Everything on this page was established the first way. ## Pine AI's twenty-six tools, and a fax machine The client exposes 26 distinct tool names. Not their schemas, only the names and how they're grouped. Sorted roughly by what they do, you get phone actions, email, web and travel search, knowledge lookup, timers and to-do management, a file parser, and two ways to interrupt you for information. And a fax tool. A personal agent shipping fax support in 2026 tells you exactly who Pine is built for. Nobody faxes their friends. You fax an insurance claims department, a medical records office, a utility that still runs on paper. That single tool name says more about the target user than the entire homepage does: it points at the places where these disputes actually get stuck. Names without schemas is its own kind of statement. A tool name tells you a capability exists. A schema tells you what it accepts, what it refuses, what it returns and what it does when something goes wrong. Pine gives you the first and withholds the second, which means the boundary of what the agent can do on your behalf is not something you can inspect. You find out by using it. ## Computer use bypasses the tool list entirely The thing the tool list doesn't cover is computer use. When Pine drives a browser, it doesn't call one of those 26 tools. The Mac app hands the instruction to a separate browser agent running on your machine, and the cloud path leases you a remote desktop you can watch over VNC and take over from. Two different execution paths, neither of them visible in the tool catalogue. This matters for anyone trying to reason about what Pine can reach. The catalogue implies a bounded set of capabilities. The browser path is not in that set and is not bounded the same way: on the Mac it runs on your own computer, outside the tool list. Whether that local browser agent can use sign-ins already saved on your machine is something we did not establish, so check it before handing Pine a task that touches a logged-in site. ## Pine AI is four products wearing one brand Pine is sold as one assistant, but it's four different request shapes underneath. There's the task assistant most people use. There's a voice copilot built on LiveKit. There's a REST API that lets other people's agents place calls through Pine. And there's the leased cloud computer. They share a name and a login. They don't share a shape, and nothing we saw suggests they share the same instructions or the same failure modes. Anyone evaluating Pine by using the task assistant is evaluating one quarter of the product. That's fine as long as you know it, and the marketing does not make it easy to know it. ## Pine AI calls from the app keep no transcript This is the finding with actual consequences for users. When Pine makes a task call on your behalf, the record it keeps afterwards is the merchant, the number, how long the call ran, what it cost you in credits, and a summary. No verbatim transcript. So when the interface marks a call as ended, that means the call stopped, not that anything was resolved, and you have no way to go back and check what was actually said on your behalf. Calls placed through Pine's developer API are different: they come back with a turn-by-turn transcript. Think about what gets committed on those calls. A cancellation. A promise to keep service through a billing date. An agreement to a retention offer at a specific price. Every one of those is a thing you might need to prove later, to the same company, and the only record is a summary written by the system that made the call. The live event stream and the saved history don't even agree on granularity. On one three-way identity-verification call we watched 62 events go by in real time; the history afterwards held 30 rows. Watch only the replay and you'll miss things that happened. Credits drift too. One voice call reported 70 credits in its final state and took 74 off the balance. Four credits is nothing. A billing system whose displayed figure and charged figure disagree at all is worth noticing, because it means one of those two numbers is not the number. ## Memory: there isn't a file you can edit Some personal agents keep your profile in plain files you can open and rewrite. Pine doesn't. There's no personality document on the client. The persona, which calls itself Mia on the phone, lives server-side, and what Pine knows about you sits in a structured fact layer that gets queried on demand rather than loaded up front. Querying on demand keeps the request small, and it scales past the point where stuffing everything into context stops working. The practical version for you is less comfortable. You can't audit or correct what it believes about you by editing a file. You go through the product or you don't go at all. If it has the wrong billing address, or the wrong account number, or an outdated note about which provider you're with, the only route to fixing it is the one the product gives you. For a product whose entire job involves speaking to companies as you, that asymmetry is the thing to weigh. ## What we couldn't establish We'd rather publish a short list of failures than a long list of guesses. - The production instruction text. Roughly 126 probes, all empty. - The schema for the calling tool. Not available from the client, not visible on the wire. - Which carrier actually places the calls. Nothing in the client, the status payloads, or the public SDK names one. - Whether Pine splits work between a fast and a slow model. Zero samples either way, so we're not going to pretend. For the user-facing jobs, see [disputing a charge](https://personalagentbench.com/tasks/dispute-a-charge/) and [cancelling a subscription](https://personalagentbench.com/tasks/cancel-a-subscription/). [Muse vs Pine](https://personalagentbench.com/muse-vs-pine/) and [Pine vs Instinct](https://personalagentbench.com/pine-vs-instinct/) separate calling, browser access and memory rather than treating the brand as one uniform capability. ## FAQ **Is Pine AI legit?** Yes, in the plain sense. Its [terms](https://www.19pine.ai/terms-of-service) name the company behind it as 19Pine Pte. Ltd., the App Store lists 19Pine Corp. as seller, and the calls it makes are real calls. The thing to weigh isn't legitimacy, it's the record: calls made from the app keep no verbatim transcript of what was said on your behalf, so anything agreed on a call exists only as a summary written by the system that made it. **Does Pine AI run on my computer?** The model does not. It runs server-side, calls are placed from Pine's servers, and the apps are displays. The one exception is browser work on the Mac app, which is handed to a separate browser agent running locally. Whether that agent can use the sign-ins already on your machine is something we did not establish. **What can Pine AI actually do?** The client exposes 26 tool names covering phone calls, email, fax, web and travel search, knowledge lookup, timers, to-dos and file parsing. Their schemas are not exposed, so the precise boundary of each capability is not something you can inspect from outside. **Can Pine AI make phone calls for me?** Yes, that is what it is built around. Calls are placed from Pine's servers, not your phone. Before a billing action it asks you to confirm the task, and when the company on the line wants the account holder, it shows a card saying what to say, which number will call and when the offer lapses. Calls made from the app leave a summary rather than a verbatim transcript. Pine also sells calls that pass a message to a friend or relative; we examined the bill-dispute product, not that one. We do not know which carrier places the calls. **Can I see what Pine AI was instructed to do?** No. The instructions never reach the client. We probed for them roughly 126 times and got nothing every time. Whether that is a deliberate choice we cannot tell from outside. **Does Pine AI record its phone calls?** It keeps the merchant, the number, the duration, the credit cost and a summary. For calls made from the app there is no verbatim transcript available to you, and the live event stream is more granular than the saved history, so replaying a call afterwards shows you less than watching it live. **How much does Pine AI cost?** Two price lists with different shapes: personal and household prices displayed per week with a monthly credit allowance, and work prices displayed per month. The quoted weekly personal price and monthly credit allowance are different units; they do not establish the billing cadence. Confirm renewal terms and any bill-negotiation success fee at checkout. The dated table in the pricing section below has the figures we read from [Pine's pricing page](https://www.19pine.ai/pricing); we keep them out of prose because vendor pricing moves. ## Where Pine AI lands on the personal agent scorecard Of the four we opened, Pine is the narrowest in what we examined and shows the least of how it works. It does one job that genuinely needed doing, it built the unglamorous tooling that job requires down to a fax endpoint, and it gives you almost no way to inspect or correct what it does in your name. That trade might be the right one for you. It's a different trade from the other three, and the [scorecard](https://personalagentbench.com/) sets all four side by side on the same six axes so you can see where they actually differ rather than where their marketing does. ## Pine AI compared with the others we opened Side by side on the same axes, against [Town](https://personalagentbench.com/pine-vs-town/), [Muse](https://personalagentbench.com/muse-vs-pine/) and [Instinct](https://personalagentbench.com/pine-vs-instinct/). We only build these for pairs we have taken apart ourselves, because a comparison with marketing on one side is the thing this site exists to replace. ## Errands we describe Pine AI against Each errand page says what the job demands and what Pine brings to it, next to the other products. They describe; they don't score this product on that errand. - [**Booking a table**](https://personalagentbench.com/tasks/book-a-table/) - [**Getting a subscription cancelled**](https://personalagentbench.com/tasks/cancel-a-subscription/) - [**Disputing a charge**](https://personalagentbench.com/tasks/dispute-a-charge/) - [**Rearranging a family week**](https://personalagentbench.com/tasks/family-scheduling/) - [**Claiming for a delayed flight**](https://personalagentbench.com/tasks/flight-delay-compensation/) - [**Watching something live**](https://personalagentbench.com/tasks/live-alerts/) - [**Organising work files**](https://personalagentbench.com/tasks/work-files/) ## What Pine AI costs Two price lists that do not share a shape. Personal and household prices are displayed per week, and work prices per month. These units do not establish renewal cadence; confirm billing cadence and fees at checkout. | Plan | Price | What it includes | | --- | --- | --- | | For personal use | $9.00/week · $6.10/week annual shown as 32% off, billed annually | 6,000 credits/month, roughly 6 to 30 tasks. up to 5 tasks in parallel. On monthly billing this tier takes a 25% success fee on bill negotiation, which the annual view drops | | For household use | $18.20/week · $12.30/week annual shown as 33% off, billed annually | 13,000 credits/month, roughly 12 to 60 tasks. up to 10 tasks in parallel | | PINE Plus | $139/month · $99.9/month annual price is per month, billed annually | 20,000 credits/month, roughly 20 to 106 tasks. batch calling up to 5 | | PINE Max | $249/month · $174.9/month annual price is per month, billed annually | 35,000 credits/month, roughly 35 to 160 tasks. batch calling up to 10 | | PINE Team | $419/month · $291.6/month annual price is per month, billed annually; marked limited time | 60,000 credits/month, roughly 60 to 310 tasks. batch calling up to 20 | Checked on the product's own pricing page, 2026-09-23. Prices move; this is what was published that day. ## Pine AI, axis by axis ## Pine AI: where it runs Server-side. Neither the web app nor the Mac app runs the model. Browser work is handed to a separate agent on your own machine; the cloud route leases you a remote desktop you watch over VNC. ## Pine AI: how tools are exposed 26 tool names reach the client, grouped by job. The schemas never do. Computer use bypasses that list entirely. ## Pine AI: where memory lives No file you can edit. A server-side persona plus a structured fact layer queried on demand. ## Pine AI: what you get to approve Confirmation is required before a billing action, and there are two separate kinds of it. A start confirmation gates the task and carries its own expiry. A different identity confirmation appears when the company on the phone wants to hear from the account holder: a card naming what to say, which number will be calling, and when the offer lapses. Both the request and the receipt land in the task history. We found no editable local memory file; deletion of server-side facts was not established. ## Pine AI: how you reach it Web and Mac, outbound phone calls, a voice copilot, and a REST endpoint other people's agents can call. Four request shapes under one brand. ## Pine AI: what we could not establish Production instruction text (roughly 126 probes, all empty), the calling tool's schema, which carrier places the calls, and whether work is split across a fast and a slow model. ## The other teardowns - [**Instinct**Lives in your messaging apps rather than in a new interface.](https://personalagentbench.com/agents/instinct/) - [**Muse**Meta's personal agent, with a Linux machine of your own behind it.](https://personalagentbench.com/agents/muse/) - [**Town**Web and native Mac, with the broadest native reach into your Mac of the four.](https://personalagentbench.com/agents/town/) Every runtime we have opened, side by side on the same axes. [See the scorecard](https://personalagentbench.com/) Or read [how we score](https://personalagentbench.com/how-scoring-works/), and [why the existing agent benchmarks](https://personalagentbench.com/benchmarks/) don't answer this. --- Published by Personal Agent Bench (https://personalagentbench.com), run by the Zentor team. Zentor builds agent software of its own, so its products are not on the scorecard and will not be added.