---
title: "Pine AI vs Instinct AI: phone-first and message-first"
description: "Pine AI vs Instinct AI: both think on a server. Pine is built around calls and borrows your browser; Instinct lives in your messages and leases one."
url: https://personalagentbench.com/pine-vs-instinct/
last_updated: 2026-10-23
---

# Pine AI vs Instinct AI, two products that look identical from one step back

Last verified **2026-10-23**

Pine AI vs Instinct AI is the pairing where the headline axis fails. Ask where each one runs and you get the same first answer twice: server-side, with the model somewhere else. On the property this bench usually leans on hardest, these two are the same product. The differences are a layer down, and they are large. One of them hands browser work to your own machine and is built around phoning companies. The other leases a browser from a third party, waits for you in WhatsApp, and has only lately added calls of its own.

- They agree on server-side work, as Town does too, and disagree about how tools and memory are exposed. Instinct's main-agent host remains unverified.
- In the routes we examined, Pine put browser work on your computer and Instinct did not, the opposite of what their front doors suggest. Instinct's desktop build was switched off on our account, so that is not the last word.
- One reaches outward to companies, the other inward to your message threads. Both could fairly be called meeting you where you are.
- Their tool surfaces sit far apart: 26 names with no schemas, against 571 actions you must read about first.
- Pine has no October completion figure, because its submissions never reached the exam in that batch. Instinct's October figure sits beside its September ones, never merged with them.

The basics first. Pine is sold at 19pine.ai, and its [App Store listing](https://apps.apple.com/us/app/pine-assistant/id6746403769) names the developer as 19Pine Corp. It [publishes its prices](https://www.19pine.ai/pricing): 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. Instinct is made by Spear Street Technology, Inc., founded by Noah Shinn. [TechCrunch reported](https://techcrunch.com/2026/09/28/viral-ai-agent-instinct-raises-1b-series-c-at-a-10b-valuation/) that it launched invite-only in August 2026, and it publishes no price. Both pricing pictures are on our teardown pages, [Pine](https://personalagentbench.com/agents/pine/#pricing) and [Instinct](https://personalagentbench.com/agents/instinct/#pricing), so they stay in one place when they change.

## The Pine and Instinct runtimes, side by side

| Axis | [Pine](https://personalagentbench.com/agents/pine/) | [Instinct](https://personalagentbench.com/agents/instinct/) |
| --- | --- | --- |
| 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. | Server orchestration with work running in a sandbox. Browser tasks use a leased cloud browser, not the Chrome on your desk. |
| How tools are exposed | 26 tool names reach the client, grouped by job. The schemas never do. Computer use bypasses that list entirely. | A command-line surface rather than a function list: 43 families and 571 actions by the product's own count, and three different totals depending on who is counting. An action refuses to run until its own help text has been read, which makes the manual part of the prompt. |
| Where memory lives | No file you can edit. A server-side persona plus a structured fact layer queried on demand. | Three stores that do different jobs: a git-tracked set of markdown memory files, an indexed layer of connector data it can query, and a fill-only vault for secrets. |
| 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. | An approval card with three answers rather than two: approve, deny, or dismiss. It carries the plan and names which device will run it. Approval binds to the exact text shown, not to the request that produced it, so a rewrite needs a new approval. Its own rules forbid an unseen assumption from riding along with an irreversible send, and it will produce a receipt on request: what it read, what it sent, what it did not touch. Deletion has a documented surface covering storage, training, location and desktop indexing. |
| 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. | Messaging first. WhatsApp, SMS and iMessage are the default surface, plus an address on its own mail domain. The web app is mostly an admin panel. |
| 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 body of the main instruction set (only its section headings surfaced), which host the main agent runs on, the production model id, and four named policy sections we found pointers to but never text for. |
| Completed, scored subset | 20% · 14 to 25 · 167 tasks | 43% · 36 to 50 · 159 tasks |

The completion row covers differing September scored subsets of adapted product runtimes using the substituted exam model, not shipping consumer products. Underlying first-batch records are unpublished; Muse's rate follows an adapter repair. [Result provenance and limitations](https://personalagentbench.com/results/). Neither figure is a ranking and we never combine them. As of 2026-09-22.

## Pine and Instinct: the same answer, and then a split

Pine 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, and the cloud route leases a remote desktop you watch over VNC.

Instinct orchestrates on a server with the work running in a sandbox. Browser tasks use a leased cloud browser, not the Chrome on your desk.

The first sentence of each is the same. The second is not, and it inverts what you would guess from how the two products present themselves. Instinct is the one that lives in your messaging apps, so it sounds closer to you. Pine is the one with the formal web interface, so it sounds further away. In the routes we examined, Pine is the product that puts an agent on your computer to do browser work, and Instinct is the one that did not: its browser work went to a leased cloud instance. Instinct also has a desktop build, which a feature flag kept shut on our account, so we cannot say what it would do on your machine.

Being easy to reach and being close to your machine are different properties, and the two products have them in opposite combinations.

## Pine vs Instinct tools: 26 names against 571 actions

Pine lets 26 tool names reach the client, grouped by job. The schemas never do, and computer use bypasses the list entirely, so the published surface does not describe what the product can actually do.

Instinct ships a command line instead of a function list: 43 families and 571 actions by its own count, with three different totals depending on who is counting. An action refuses to run until its own help text has been read, which makes the documentation part of the prompt and part of the bill.

The two sit far apart, and they differ in kind as well as in size. Pine is sparing about what it tells the client and capable of browser work through computer use, with limits we did not establish. Instinct lists a great deal and is strict about the conditions. One is a small door to a room we could not fully measure. The other is a large door with a reading requirement at the threshold.

The practical consequence is about predictability. With Instinct you can enumerate the action surface and know the agent has loaded a description before acting. For Pine we enumerated 26 tool names, but schemas and the separate computer-use surface remain incomplete; the list does not establish the full capability boundary.

It also changes where you would look for a mistake. Instinct's help text is inside the loop, so a stale description could produce a wrong action that looks like a model error. Pine sends no schemas to the client, and a wrong action through computer use would not trace back to any listed tool. We did not test either failure; the point is that you would hunt for them in different places.

## Pine and Instinct on calls and messages

Pine phones companies about bills, refunds and cancellations. Four request shapes sit under that one brand: the web and Mac apps, outbound calls, a voice copilot, and a REST endpoint other software can call.

Instinct is messaging first. WhatsApp, SMS and iMessage are the default surface, plus an address on a mail domain it provides. The web app is mostly an admin panel, and on the account we examined its chat route was switched off. It can make calls too now: [TechCrunch reported](https://techcrunch.com/2026/09/28/viral-ai-agent-instinct-raises-1b-series-c-at-a-10b-valuation/) a concierge feature that phones businesses for you, for things like appointments at places without online booking. We have not examined that feature. Pine is built around the call; for Instinct it is one feature added to a messaging product.

Both could be described as meeting you where you already are, and the phrase points in two different directions. Instinct's version is that you never open anything: the agent is in the thread you were already in. Pine's version is that the errand reaches the company without you being in the conversation at all.

That second one is the more unusual capability and the less examined. A call puts an agent in front of a person who did not agree to talk to a machine, about your account. When a call from Pine's app goes wrong, what you have is a summary written by the system that made it; calls through its developer API at least return a transcript.

Calls, messages and bookings can all have consequences that software cannot fully undo. Deleting a message does not undo it being read, and cancellation can carry costs.

## Instinct and Pine on memory: three stores against a fact service

Instinct keeps three stores doing three jobs: a git-tracked set of markdown memory files, an indexed layer of connector data it can query, and a fill-only vault for secrets built to accept credentials and never return them.

Pine has no file you can edit: a server-side persona and a structured fact layer queried on demand.

So Pine gives you nothing to open, while Instinct keeps markdown files under version control, though the product decides what goes into them and we did not establish whether you can edit them yourself. The difference between these two is what kind of thing is holding the state. Instinct has objects: a set of files, an indexed store, a vault, each with its own rules, and you can at least reason about which is which. Pine has a service that answers questions about you, and reasoning about it means reasoning about query results.

If the product has the wrong idea about you, Instinct gives you a place to point at and Pine gives you a behaviour to complain about.

## What wakes Pine and Instinct

Instinct can be started by a clock or by an event on your mail, calendar, documents or location, with a platform-side check before a wake becomes an action. You are not the only thing that wakes it up.

Pine's REST endpoint means other people's software can drive it, which is a related property arriving from a different direction: the request is not necessarily typed by you. We did not establish whether Pine starts work on its own schedule.

Put that next to the browser arrangement and Pine's shape comes into focus. Software other than you can start it, it can place a telephone call, and when it needs a browser it may use an agent on your own machine. Each of those is defensible alone. Together they describe a product with more reach into the world than its quiet web interface implies.

## Pine vs Instinct: which one asks before it acts

Both ask before the consequential step, and both show you what they are asking about. What they gate is different.

Pine gates the task and the identity. A start confirmation gates the task and carries its own expiry. A separate 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. [Pine's control row](https://personalagentbench.com/agents/pine/#control) has the detail.

Instinct gates each action. In the account we examined, an approval card carries the plan, names the device that will run it, and offers approve, deny or dismiss. Approval binds to the exact text shown, so a rewrite needs a fresh approval. Its own rules forbid an unseen assumption from riding along with an irreversible send, and it will give a receipt on request listing what it read, what it sent and what it did not touch. The card covers actions, not reading: in one [reported case](https://personalagentbench.com/is-instinct-safe/#instinct-login-codes) it found a login code in a connected inbox and used it without asking.

So the moment to pay attention differs. With Pine it is the start of the task and any point where a company asks for you. With Instinct it is every card, because the exact text on it is what you approve.

## The same wall in the Pine and Instinct teardowns

For Pine we could not get the production instruction text, after roughly 126 probes that all came back empty, nor the calling tool's schema, nor which carrier places the calls, nor whether the work is split between a fast and a slow model.

For Instinct we could not get the body of the main instruction set, only its section headings, nor which host the main agent runs on, nor the production model id, nor four named policy sections we found pointers to but no text for.

Both lists stop in the same place. We can describe the machinery around the instructions in detail and we cannot read the instructions. That is the boundary of examination from outside, and the honest thing a comparison can do is mark it rather than write past it.

The Pine gap worth repeating is the carrier. The product's distinguishing feature is a phone call, and we do not know whose network carries it. That is not an exotic detail. Whoever carries the call is a party to every conversation the product has on your behalf, and the product does not name them.

## Pine vs Instinct: which one the choice comes down to

One question separates them, and it is not about capability.

Does your errand end with a company, or with you? Cancellations, refunds and billing disputes end with a company, often on the phone, and Pine is built end to end for that. Scheduling, triage, following something up and keeping track end with you, and Instinct is built to live in the place where that conversation already happens.

The second question is how much of the surface you want listed. Instinct lists its actions and makes the agent read about each one first. Pine sends 26 names without schemas, and its computer-use route is not in that list, so the list does not mark the edge of what it can do. If you want to reason about what the thing might do, that difference matters more than any number on our scorecard.

Put as fit rather than as a ranking:

- **Pick Pine if** the errand ends with a company, a cancellation, a refund or a billing dispute, and you want to confirm at the start and again whenever the company asks for the account holder.
- **Pick Instinct if** the errand ends with you, scheduling, triage or following something up, and you want it handled from WhatsApp, SMS or iMessage with an approval card that shows the exact text of the action it proposes.

## What the figures say

Both carry a September completion rate with an interval and a task count, shown in the table near the top of this page. Instinct's interval (36 to 50%) sits above Pine's (14 to 25%), but the two rest on different draws of tasks, so the gap is not a head-to-head result.

It is still not a ranking, and this pairing is where that caveat bites hardest. The scored set mixes errands that end with a company and errands that end with you, and these two products are built for different halves of it. A single rate across the whole set is a statement about the mix as much as about the products. Why we print rates only with their intervals, and never as one blended score, is in [how scoring works](https://personalagentbench.com/how-scoring-works/); the same Instinct figure sits against Muse's in [Instinct vs Muse](https://personalagentbench.com/muse-vs-instinct/).

On the separate [114-task batch](https://personalagentbench.com/results/), measured on 22 September 2026, Pine completed 30% (95% interval 23 to 36%; 546 attempts) and Instinct completed 52% (95% interval 45 to 61%; 380 attempts). Both answered the same 114 tasks. These are replicas of older runtimes with the same GLM 5.3 flash model substituted, under older rules, filtered from roughly 5,600 attempts using eight rules, not fresh product results. Excluding attempts that never reached the exam removes over 400 each for Pine and Muse; if those were refusals rather than broken entry, the supplied alternative estimates are about 22% for Pine and 24% for Muse. Transcripts were supplied but have not been republished here; some files were overwritten or shared between attempts, so the underlying record is incomplete. Keep this measurement separate from the main table and do not treat interval overlap or a point estimate as a significance test.

The October batch, run on 11 October 2026, gave every product the same 20 tasks once each, under the exam rules that took effect at the end of September. Instinct completed 6 of 15 scored attempts, 40% (95% interval 20 to 64%). Pine has no October figure: almost none of its submissions reached the exam, so its column was set aside rather than scored as zero, and the cause is not yet known and may be on the exam side. The tasks, the rules and our replicas all changed between September and October, so a different rate is not a product getting better or worse, and the batches are never added together. The other October batches, each with its own tasks and builds, are listed separately on [the results page](https://personalagentbench.com/results/), and individual recorded attempts are on [the run pages](https://personalagentbench.com/runs/), as counts rather than rates.

The [Pine pricing snapshot](https://personalagentbench.com/agents/pine/#pricing) starts personal use at $9.00/week, dated 2026-09-23. Weekly personal pricing is not directly comparable to a monthly work plan; the page separates billing periods and additional fees.

## FAQ

**What is the main difference between Pine and Instinct?** What each one reaches for. Pine is built around phoning companies about bills, refunds and cancellations. Instinct lives in your messaging apps and works from the thread you are already in, and has added a concierge feature that calls businesses, which we have not examined.

**Do either of them run on my computer?** Neither runs the model locally. Pine hands browser work to a separate agent on your own machine and leases a remote desktop on its cloud route. Instinct used a leased cloud browser in the examined account; its desktop build was switched off by a feature flag there, so we never saw it run.

**Which exposes more tools?** Instinct, by a wide margin on paper: 571 actions across 43 families, with the help text required before an action runs. Pine sends 26 names and no schemas, and its computer use goes around that list, so the list does not show its whole surface.

**Can I correct what either one knows about me?** Pine gives you no file: a server-side persona and a fact layer queried on demand. Instinct keeps a git-tracked set of markdown memory files, an indexed layer of connector data and a fill-only vault. A corrected fact stays in the files' history, but we did not establish whether you can edit them yourself.

**Can something other than me start them?** Instinct can be woken by a clock or an event on your mail, calendar, documents or location. Pine can be driven by other software through its REST endpoint; we did not establish whether it starts work on its own.

**Who makes Pine and Instinct, and what do they cost?** Pine's terms name 19Pine Pte. Ltd.; the App Store seller is 19Pine Corp. It publishes weekly personal plans and monthly work plans. Instinct is made by Spear Street Technology, launched invite-only in August 2026, according to TechCrunch, and publishes no price. Details are on our [Pine](https://personalagentbench.com/agents/pine/#pricing) and [Instinct](https://personalagentbench.com/agents/instinct/#pricing) pages.

**Is this from testing or from their documentation?** From our own teardowns of [Pine](https://personalagentbench.com/agents/pine/) and [Instinct](https://personalagentbench.com/agents/instinct/), including the list of what we could not establish for each.

## Other comparisons

- [**Muse vs Instinct**Compared on what each one actually runs · Instinct AI vs Muse, from our teardowns of both: where each runs, how you reach it, which asks before acting, phone calls, and which fits your errand.](https://personalagentbench.com/muse-vs-instinct/)
- [**Pine vs Town**A borrowed browser or a Mac foothold · Pine AI vs Town AI: Pine thinks on its servers and borrows a browser; Town's Mac app reaches your messages, contacts, screen and audio. From our teardowns.](https://personalagentbench.com/pine-vs-town/)
- [**Instinct vs Town**Invite, and who asks first · Instinct AI vs Town AI: Instinct is invite-only and publishes no price; Town publishes prices, starting with a free plan. Instinct asks on a card; Town asks in the routine's settings.](https://personalagentbench.com/instinct-vs-town/)
- [**Muse vs Pine**And what a machine of your own is actually for · Meta Muse vs Pine AI: Muse gives every user a Linux VM in Meta's cloud; Pine runs its model on its servers and borrows a browser. From our own teardowns.](https://personalagentbench.com/muse-vs-pine/)
- [**Muse vs Town**And two things people call a machine · Meta Muse vs Town AI: Muse gives each user a Linux VM in Meta's cloud; Town's native Mac app reaches your messages and screen. From our own teardowns.](https://personalagentbench.com/muse-vs-town/)

All six pairs, and why there are only six, are on [the comparisons page](https://personalagentbench.com/compare/).

## More on Instinct

- [**Instinct teardown**](https://personalagentbench.com/agents/instinct/)
- [**How Instinct AI works**How does Instinct AI work? One errand traced from trigger to reply: what wakes it, how it picks tools, where it runs, what it remembers and when it asks.](https://personalagentbench.com/how-instinct-ai-works/)
- [**Instinct AI review**Instinct AI review roundup: what reporters who used it got done and got wrong, how that maps onto our teardown, and our exam figures with intervals.](https://personalagentbench.com/instinct-ai-review/)
- [**Is Instinct safe?**Is Instinct AI safe? Every incident users and reporters have described, from unchecked emails to login codes read from Gmail, checked against our teardown.](https://personalagentbench.com/is-instinct-safe/)

---

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.
