---
title: "Personal AI agents list: products and evidence"
description: "Personal AI agents list: what each product does, who makes it, how to reach it, and whether we tested it or read published sources."
url: https://personalagentbench.com/landscape/
last_updated: 2026-11-05
---

# Personal AI agents list: products and evidence

This personal AI agents list records assistants that act on your accounts rather than only answering questions. New products arrive faster than we can examine them. This page is the list we keep so that the gap between what exists and what we have examined is visible rather than implied. Four of them have been through the exam, and a fifth, [Grok Bot](https://personalagentbench.com/grok-bot/), is listed in the exam data as pending, without a published figure. Separately, [15 open-source agents](https://personalagentbench.com/open-source-personal-agents/), including [OpenMuse](https://personalagentbench.com/openmuse/), have pinned source reviews without runtime testing. The rest have source links; several also have public-record pages based on published sources, which are distinct from runtime testing. If you want measured results, they are on [the scorecard](https://personalagentbench.com/), with their intervals. Skip to the list.

- The list separates what we have examined from what merely exists. Most entries are in the second group and say so.
- Being on this list is not an endorsement and the order is not a ranking. Frameworks, source templates and memory components form a separate group; within each group, entries are ordered by our coverage, not product quality.
- Every entry carries a source link you can open. We label a source type where established; some rows retain the generic label "Source link". Names we could not trace to an official site or a first-hand report are held back, and the note under the table says how many.
- Most rows carry a date from last week or this one, because a job checks them every Monday. The handful nothing can check automatically say so instead of sitting there looking current.
- There is no score column. In the September first batch, three products have figures and Town's is withheld. In the 11 October batch, where all four answered the same 20 tasks, Instinct, Muse and Town have figures and Pine was set aside with no data. The older common-task batch includes a qualified Town rate. None of those figures transfers to anything else on this page.

## The list

| Product | How you reach it | Purpose and access | Where we have got to | Source |
| --- | --- | --- | --- | --- |
| Assistants and task products |   |   |   |   |
| [Instinct](https://personalagentbench.com/agents/instinct/) · Spear Street Technology | A message thread | Messaging assistant for delegated tasks Examined account; current access may differ | **Scored**. Through the exam, rate on the scorecard | [Source link](https://instinct.com/) Not watched automatically |
| [Muse](https://personalagentbench.com/agents/muse/) · Meta | Phone apps and web | Personal assistant with a per-user VM Examined account; current access may differ | **Scored**. Through the exam, rate on the scorecard | [Source link](https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/) Checked weekly, last 2026-11-09 |
| [Pine](https://personalagentbench.com/agents/pine/) · 19pine | It telephones for you | Business calls, bills and cancellations Examined account; current access may differ | **Scored**. Through the exam, rate on the scorecard | [Source link](https://www.19pine.ai/pricing) Checked weekly, last 2026-11-09 |
| [Town](https://personalagentbench.com/agents/town/) · town.com | Web and a Mac app | Context, briefings and routines Examined account; current access may differ | **Scored**. First-batch figure withheld; qualified older common-task rate on the scorecard | [Source link](https://www.town.com/pricing) Checked weekly, last 2026-11-09 |
| Agibot · Ökte Çilmi | A desktop app | Specific task scope not independently established Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Product Hunt listing](https://www.producthunt.com/products/agibot) Checked weekly, last 2026-11-09 |
| Aico Link (formerly Aiden) · AidenAI-IO | Its own hardware | Hardware add-on that operates a phone through its screen; in development Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://aico.link/) Checked weekly, last 2026-11-09 |
| Alexa+ · Amazon | Phone apps and web | Specific task scope not independently established Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://aboutamazon.com/news/devices/alexa-plus-web-ai-assistant/) Not watched automatically |
| Asaply · Asaply | A message thread | Text-based everyday errands Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://asaply.ai/) Checked weekly, last 2026-11-09 |
| Boba · humans& | A message thread | Text and group-chat assistant for goals and plans Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://boba.app/) Checked weekly, last 2026-11-09 |
| Boski · Boski | A message thread | Subscription cancellations, bill negotiation and refund claims, as its site describes Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://www.boski.com/) Checked weekly, last 2026-11-09 |
| Brea · Brea | A message thread | Tasks across existing apps Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://brea.ai/) Checked weekly, last 2026-11-09 |
| [Claude Cowork (Dispatch)](https://personalagentbench.com/claude-dispatch/) · Anthropic | A desktop app | Phone-directed desktop work Existing users only in the cited October help page | **Listed**. Recorded that it exists, nothing more | [Source link](https://support.claude.com/en/articles/13947068-assign-tasks-to-claude-from-anywhere-in-cowork) Checked weekly, last 2026-11-09 |
| [Claude Duet](https://personalagentbench.com/claude-duet/) · Anthropic | Phone apps and web | Proactive agent on a cloud computer US Pro and Max waitlist, research preview | **Listed**. Recorded that it exists, nothing more | [Source link](https://claude.com/resources/articles/claude-duet) Checked weekly, last 2026-11-09 |
| Comma · AFK AI, Inc. | You run it yourself | Self-hosted always-on assistant with browser and computer use Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://comma.surf/) Checked weekly, last 2026-11-09 |
| [Copilot Tasks](https://personalagentbench.com/copilot-tasks/) · Microsoft | Phone apps and web | Task product migrating to Cowork Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Official help page](https://support.microsoft.com/en-us/microsoft-copilot/using-copilot-tasks) Checked weekly, last 2026-11-09 |
| [Cue](https://personalagentbench.com/cue-by-manus/) · Manus (Butterfly Effect) | Phone apps and web | Personal team of persistent agents Invite-gated early access; code availability unverified | **Listed**. Recorded that it exists, nothing more | [Source link](https://manus.im/blog/introducing-manus-2-0) Checked weekly, last 2026-11-09 |
| [Dot](https://personalagentbench.com/openai-dots/) · OpenAI | Inside ChatGPT | Persistent agents inside ChatGPT Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Third-party comparison](https://every.to/personal-agents-comparison) Not watched automatically |
| Dot (Emotion Computer) · Emotion Computer | On your phone | Separate personal-assistant app Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://www.emotioncomputer.com/) Checked weekly, last 2026-11-09 |
| Duckbill · Duckbill | Phone apps and web | Specific task scope not independently established Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://www.getduckbill.com/) Checked weekly, last 2026-11-09 |
| Folk · Nozomio Labs | A message thread | Coaching and personal admin Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://www.folk.com/) Checked weekly, last 2026-11-09 |
| Gemini Spark · Google | Phone apps and web | Ongoing tasks across Google apps Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Third-party comparison](https://every.to/personal-agents-comparison) Not watched automatically |
| [Grok Bot](https://personalagentbench.com/grok-bot/) · X Corp. (App Store), SpaceXAI (Google Play) | A message thread | Persistent bots for work across apps Current account access not independently checked | **Listed; exam pending**. In the exam data, no published figure or runtime teardown | [App Store listing](https://apps.apple.com/us/app/grok-bot/id6794501026) Checked weekly, last 2026-11-09 |
| Halo · Tanmay Sonawane | On your phone | Phone-based personal assistant Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://meethalo.app/) Checked weekly, last 2026-11-09 |
| Hark · Hark Labs Inc. | Phone apps and web | Specific task scope not independently established Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://hark.com/) Checked weekly, last 2026-11-09 |
| [Hermes Agent](https://personalagentbench.com/hermes-agent/) · Nous Research | You run it yourself | Self-hosted assistant with memory and skills Current account access not independently checked | **Read from source**. Pinned source reviewed with REA where applicable and manual reading; not run or scored | [Source link](https://hermes-agent.nousresearch.com/) Checked weekly, last 2026-11-09 |
| Inbox Zero · Elie Steinbock | Email | Specific task scope not independently established Repository available; deployment untested | **Listed**. Recorded that it exists, nothing more | [Repository](https://github.com/elie222/inbox-zero) Checked weekly, last 2026-11-09 |
| Jarvis AI Assistant · Adrees Umer | A desktop app | Specific task scope not independently established Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://jarvis.institute/) Checked weekly, last 2026-11-09 |
| [Leon](https://personalagentbench.com/open-source-personal-agents/#leon) · Open source | You run it yourself | Specific task scope not independently established Repository available; deployment untested | **Read from source**. Pinned source reviewed with REA where applicable and manual reading; not run or scored | [Repository](https://github.com/leon-ai/leon) Checked weekly, last 2026-11-09 |
| Lindy · Lindy | A message thread | Specific task scope not independently established Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://www.lindy.ai/) Checked weekly, last 2026-11-09 |
| [LobsterAI](https://personalagentbench.com/open-source-personal-agents/#lobsterai) · NetEase Youdao | A desktop app | Specific task scope not independently established Repository available; deployment untested | **Read from source**. Pinned source reviewed with REA where applicable and manual reading; not run or scored | [Repository](https://github.com/netease-youdao/LobsterAI) Checked weekly, last 2026-11-09 |
| Martin · DCWU | A message thread | Research, reminders and calendar tasks Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://trymartin.com/) Checked weekly, last 2026-11-09 |
| Maverick · S.R.L. Maverick One | A message thread | Bookings, cancellations and plans in iMessage, with approval before acting Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://usemaverick.io/) Checked weekly, last 2026-11-09 |
| Microsoft Autopilot (formerly Scout) · Microsoft | Phone apps and web | Always-on work agent in Microsoft 365, built from OpenClaw Private preview for organizations, per Microsoft's 25 September post | **Listed**. Recorded that it exists, nothing more | [Official blog post](https://blogs.microsoft.com/blog/2026/09/25/introducing-the-new-copilot-with-home-code-and-autopilot/) Checked weekly, last 2026-11-09 |
| [MimiClaw](https://personalagentbench.com/open-source-personal-agents/#mimiclaw) · Memov AI | Its own hardware | Specific task scope not independently established Repository available; deployment untested | **Read from source**. C firmware read manually; REA not applicable; not run or scored | [Repository](https://github.com/memovai/mimiclaw) Checked weekly, last 2026-11-09 |
| [nanobot](https://personalagentbench.com/open-source-personal-agents/#nanobot) · HKUDS | You run it yourself | Specific task scope not independently established Repository available; deployment untested | **Read from source**. Pinned source reviewed with REA where applicable and manual reading; not run or scored | [Repository](https://github.com/HKUDS/nanobot) Checked weekly, last 2026-11-09 |
| [NanoClaw](https://personalagentbench.com/open-source-personal-agents/#nanoclaw) · nanocoai | You run it yourself | Specific task scope not independently established Repository available; deployment untested | **Read from source**. Pinned source reviewed with REA where applicable and manual reading; not run or scored | [Repository](https://github.com/nanocoai/nanoclaw) Checked weekly, last 2026-11-09 |
| [nanoMuse](https://personalagentbench.com/nanomuse/) · nanoMuse contributors (Zhejiang University) | You run it yourself | Device-hosted personal assistant Repository available; deployment untested | **Read from source**. Pinned source reviewed; live community relay settings not established | [Repository](https://github.com/nano-muse/nanoMuse) Checked weekly, last 2026-11-09 |
| Ollie · Ollie | A message thread | Email, calendar and reminders Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://try.ollie.ai/) Checked weekly, last 2026-11-09 |
| OpenCall · Calify | It telephones for you | Outbound calling assistant Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://www.calify-ai.com/) Checked weekly, last 2026-11-09 |
| [OpenClaw](https://personalagentbench.com/openclaw-security/) · Open source | You run it yourself | Self-hosted assistant with messaging channels Current account access not independently checked | **Read from source**. Pinned source reviewed with REA where applicable and manual reading; not run or scored | [Source link](https://docs.openclaw.ai/) Checked weekly, last 2026-11-09 |
| [OpenInstinct](https://personalagentbench.com/open-source-personal-agents/#openinstinct) · Merit Systems | You run it yourself | Community messaging assistant with a vault Repository available; deployment untested | **Read from source**. Pinned source reviewed with REA where applicable and manual reading; not run or scored | [Repository](https://github.com/Merit-Systems/OpenInstinct) Checked weekly, last 2026-11-09 |
| OpenMausBot · Milind Soni (open source) | A desktop app | Open-source chat app that runs a team of agent bots on your own machine Repository available; deployment untested | **Listed**. Recorded that it exists, nothing more | [Repository](https://github.com/milind-soni/OpenMausBot) Checked weekly, last 2026-11-09 |
| [OpenPoke](https://personalagentbench.com/open-source-personal-agents/#openpoke) · Community implementation | You run it yourself | Community implementation for messaging agents Repository available; deployment untested | **Read from source**. Pinned source reviewed with REA where applicable and manual reading; not run or scored | [Repository](https://github.com/shlokkhemani/openpoke) Checked weekly, last 2026-11-09 |
| Orbits · NemusAI, Inc. | Phone apps and web | Household calendars, bookings, purchases and phone calls Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://tryorbits.com/) Checked weekly, last 2026-11-09 |
| Orchid · Orchid | A message thread | Messaging assistant and automation Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://orchid.ai/) Checked weekly, last 2026-11-09 |
| Pally · Pally | A message thread | Messaging assistant for bookings and purchases Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://pally.com/) Checked weekly, last 2026-11-09 |
| Partyline · Partyline | A message thread | Specific task scope not independently established Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://partyline.space/) Checked weekly, last 2026-11-09 |
| Perplexity Personal Computer · Perplexity | Web and a Mac app | Specific task scope not independently established Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [News report](https://engadget.com/ai/perplexity-brings-its-personal-computer-ai-assistant-to-mac-202045969.html) Not watched automatically |
| Pickle · Pickle | A computer of its own | Bookings, orders and payments on an agent computer Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://pickle.com/) Checked weekly, last 2026-11-09 |
| [Poke](https://personalagentbench.com/poke-ai/) · The Interaction Company, acquired by Cognition in July 2026 | A message thread | Messaging assistant and human task handoff Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://poke.com/) Checked weekly, last 2026-11-09 |
| Rene · Second Enlightenment Ltd | A message thread | Browser tasks, bookings and scheduling by text Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://rene.co/) Checked weekly, last 2026-11-09 |
| szn · szn | A message thread | Calls, email and follow-ups for everyday tasks Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://theszn.ai/) Check failed 2026-11-09 |
| Tab · Terrasoft, Inc. | A message thread | Research, scheduling, calls and purchases by text Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://tab.bot/) Checked weekly, last 2026-11-09 |
| Today · Today AI | Phone apps and web | Specific task scope not independently established Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [App Store listing](https://apps.apple.com/us/app/today-personal-ai-assistant/id6760608312) Checked weekly, last 2026-11-09 |
| Tomo · Tomo | A message thread | Goal-oriented messaging assistant Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://www.tomo.ai/) Checked weekly, last 2026-11-09 |
| Underdog · Conway Research | A desktop app | Email, calendar and errands on a local model Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://underdog.ai/) Checked weekly, last 2026-11-09 |
| Vellum · Vellum | Phone apps and web | Specific task scope not independently established Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://www.vellum.ai/) Checked weekly, last 2026-11-09 |
| Wabi · Wabi, Inc. | Phone apps and web | Group-chat assistant for appointments, returns and follow-ups Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [App Store listing](https://apps.apple.com/us/app/wabi-ai-assistant/id6747768928) Checked weekly, last 2026-11-09 |
| Wajo · Wajo | Phone calls and email | Calls, email and coordination with human support Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://wajo.ai/) Checked weekly, last 2026-11-09 |
| Zapia · BrainLogic AI | A message thread | Business enquiries and bookings Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://zapia.com/) Checked weekly, last 2026-11-09 |
| [zeroclaw](https://personalagentbench.com/zeroclaw/) · zeroclaw labs | You run it yourself | Specific task scope not independently established Repository available; deployment untested | **Read from source**. Pinned source reviewed with REA where applicable and manual reading; not run or scored | [Repository](https://github.com/zeroclaw-labs/zeroclaw) Checked weekly, last 2026-11-09 |
| Zo Computer · Zo | A computer of its own | Specific task scope not independently established Current account access not independently checked | **Listed**. Recorded that it exists, nothing more | [Source link](https://www.zo.computer/) Checked weekly, last 2026-11-09 |
| Frameworks, source templates and memory components |   |   |   |   |
| [OpenMuse](https://personalagentbench.com/openmuse/) · CopilotKit | You run it yourself | Source template for a personal agent Repository available; deployment untested | **Read from source**. Source code read, not probed or scored | [Repository](https://github.com/CopilotKit/openmuse) Checked weekly, last 2026-11-09 |
| [CowAgent](https://personalagentbench.com/open-source-personal-agents/#cowagent) · Open source | You run it yourself | Agent framework and task harness Repository available; deployment untested | **Read from source**. Pinned source reviewed with REA where applicable and manual reading; not run or scored | [Repository](https://github.com/zhayujie/CowAgent) Checked weekly, last 2026-11-09 |
| [MIRIX](https://personalagentbench.com/open-source-personal-agents/#mirix) · Mirix AI | You run it yourself | Memory infrastructure; not equivalent to a consumer agent Memory project; consumer-product availability unverified | **Read from source**. Pinned source reviewed with REA where applicable and manual reading; not run or scored | [Repository](https://github.com/Mirix-AI/MIRIX) Checked weekly, last 2026-11-09 |
| [QwenPaw](https://personalagentbench.com/open-source-personal-agents/#qwenpaw) · AgentScope | You run it yourself | Local or cloud agent framework Repository available; deployment untested | **Read from source**. Pinned source reviewed with REA where applicable and manual reading; not run or scored | [Repository](https://github.com/agentscope-ai/QwenPaw) Checked weekly, last 2026-11-09 |

Checked 2026-10-09. 8 further names appear on a competitor's public list and are held back here, because we could not trace them to an official site or a first-hand report.

The part worth knowing about is the last column. A list like this is usually a snapshot that quietly rots, so most of these entries are re-checked every Monday by a job that looks at the App Store, the repository or the vendor's own page and reports what moved. Each row says when it was last looked at. A few cannot be checked that way, and those rows say that rather than borrowing the freshness of the ones that can.

| Product | What the cited coverage describes | Evidence |
| --- | --- | --- |
| Pine | Calls to companies, browser work | [Teardown](https://personalagentbench.com/agents/pine/) |
| Instinct | Messaging, delegated work and account actions | [Teardown](https://personalagentbench.com/agents/instinct/) |
| Muse | Per-user VM, connected services, Mac app | [Teardown](https://personalagentbench.com/agents/muse/) |
| Town | Briefings, inbox work and routines | [Teardown](https://personalagentbench.com/agents/town/) |
| Dots | Background and scheduled work in ChatGPT | [Public record](https://personalagentbench.com/openai-dots/) |
| Claude Duet | Proactive agent on a cloud computer, waitlist | [Public record](https://personalagentbench.com/claude-duet/) |
| Dispatch | Tasks assigned from a phone to Cowork | [Public record](https://personalagentbench.com/claude-dispatch/) |
| Copilot Tasks | One-off and scheduled tasks; Cowork migration | [Public record](https://personalagentbench.com/copilot-tasks/) |
| Grok Bot | Named bots using cloud computers | [Public record](https://personalagentbench.com/grok-bot/) |
| Poke | Messaging, reminders and connected accounts | [Public record](https://personalagentbench.com/poke-ai/) |
| Cue | App-based agent and cloud computer | [Public record](https://personalagentbench.com/cue-by-manus/) |
| nanoMuse | Device agents joined by a relay | [Public record](https://personalagentbench.com/nanomuse/); [source review](https://personalagentbench.com/open-source-personal-agents/#nanomuse) |

## What the three states mean

**Scored** means the product has been through the exam. Four products are in this state, and which of them has a figure depends on the batch. In the September first batch, Town's figure is withheld because it cannot be reconciled with the available records. In the 11 October batch on [the scorecard](https://personalagentbench.com/), Pine was set aside because almost none of its submissions reached the exam. Every published figure has an interval and a task count beside it. A separate older common-task batch includes a Town rate with its own provenance limits.

**Read from source** means we have examined how it is built and published what we found, without it having been through the exam. That answers a different question from a rate: one is about construction, the other about outcome. The [15-agent source comparison](https://personalagentbench.com/open-source-personal-agents/) extends the approach used for [OpenMuse](https://personalagentbench.com/openmuse/): reading code at a fixed version without running the agents. On JavaScript and TypeScript projects a mapping tool traced imports, storage and outgoing requests; conclusions about behaviour, such as approval paths, defaults and telemetry, were confirmed by reading the source.

**Listed** means we have recorded that it exists, what it is reached through, and where to read about it. Some listed products also have public-record pages. That coverage does not imply a teardown or an exam score.

The states are about our coverage, not about product quality. A listed product is not worse than a scored one. It is one we have not got to.

## Why most personal AI agents here are only listed

Taking a product apart is slow. Getting an account, watching what it does, working out where the model runs and what it can reach, and then writing only what the evidence supports takes far longer than a new one takes to launch.

That asymmetry is permanent, and pretending otherwise would mean either thinning the work until it says nothing, or quietly dropping the products we have not reached. This page is the third option. It keeps the unexamined ones visible, with their state attached, so the shape of what we have not done stays legible.

A list that only contained what we had examined would look more complete and be less true.

## What personal AI agents are, and what we record about each

A personal AI agent is software aimed at one person that acts for them rather than only answering: it reaches their accounts, keeps state about them between conversations, and runs errands such as booking, cancelling, disputing a bill or keeping a calendar. The directory includes consumer agents alongside related workplace products, frameworks and memory components. A listing is discovery evidence, not a claim that every entry is a shipping consumer assistant.

Underneath, every one we have opened splits into the same few parts, and they differ far more there than in their feature lists:

- **Where the model runs and what it can touch.** None of the four we opened runs the model on your machine. [Pine](https://personalagentbench.com/agents/pine/) runs the model and places its calls from its own servers, [Muse](https://personalagentbench.com/agents/muse/) gives each user a Linux VM, and [Town](https://personalagentbench.com/agents/town/) puts a native Mac app with system access on your laptop.
- **How tools reach the model.** Pine shows 26 tool names and no schemas. Town narrows a catalogue of 1105 to 177 for its main assistant.
- **Where memory lives.** Muse keeps plain markdown on a per-user VM; personal access to those files was not verified. Pine keeps a server-side fact layer whose direct inspection, editing and deletion controls we did not establish.
- **What it asks before acting, and how you reach it.** The second is the surface column in the list above.

[How we take them apart](https://personalagentbench.com/how-we-take-them-apart/) explains how those runtime findings were established; the [open-source reviews](https://personalagentbench.com/open-source-personal-agents/) use a separate static method.

For everything else we record four fields, deliberately few: the **name and who makes it**, because ownership changes; the **surface**; the **state**; and a **link** you can open. No pricing, no ratings, no feature grid. Pricing for the examined products sits on their own pages with the date it was checked, because prices go stale faster than anything else on a site like this.

## What a personal AI assistant's surfaces tell you

Sort this list by how you reach each product and the category splits into groups that behave differently, which is why that column is there rather than a feature count.

The largest group lives in a message thread: Instinct, [Poke](https://personalagentbench.com/poke-ai/), [Grok Bot](https://personalagentbench.com/grok-bot/) and Lindy among them. You text it the way you would text a person, and it answers in the same place. Approval depends on the product: Instinct has an observed card with approve, deny and dismiss, bound to exact plan text. A messaging surface does not establish whether an action can bypass review. Cards and other controls depend on the particular messaging interface. It also means the agent can reach you whenever it decides to, which is the same capability seen from the other side.

A smaller group is a web product with apps attached: Muse, [Cue](https://personalagentbench.com/cue-by-manus/), Microsoft's [Copilot Tasks](https://personalagentbench.com/copilot-tasks/) and Amazon's Alexa+. There is a surface to put a decision on, which matters whenever the next step costs money or sends something irreversible, and a place for a notification to sit rather than having to interrupt you.

A few telephone on your behalf, Pine and OpenCall among them. That is a narrower job and a harder one, because a phone call cannot be undone, paused or edited once the other party has heard it.

Some sit on your desktop instead, such as Anthropic's [Claude Dispatch](https://personalagentbench.com/claude-dispatch/), which its [help page](https://support.claude.com/en/articles/13947068-assign-tasks-to-claude-from-anywhere-in-cowork) says works on your computer while you message it from your phone.

And several you run yourself: [OpenClaw](https://personalagentbench.com/openclaw-security/), its Rust reimplementation [ZeroClaw](https://personalagentbench.com/zeroclaw/), Nous Research's [Hermes Agent](https://personalagentbench.com/hermes-agent/) and [OpenMuse](https://personalagentbench.com/openmuse/) among them. Those trade convenience for the ability to read the source, which is a real trade rather than a marketing distinction: it permits source inspection, while hosted dependencies and actual deployments still need separate verification. Self-hosting does not guarantee local-only data or disabled telemetry; the [source comparison](https://personalagentbench.com/open-source-personal-agents/) keeps those questions separate.

A last group works a phone screen instead of a chat thread or a browser. The open-source ones are compared on [phone use agents](https://personalagentbench.com/phone-use/), by how each one reaches the screen and what that lets it touch.

The surface shapes interaction, but it does not by itself determine approvals, memory or reach. That is why it is the first question every teardown on this site asks.

The surface also decides how a product can reach you first. A message thread can open a conversation, a web product needs a notification to do it, and [who acts unasked](https://personalagentbench.com/proactive-ai-agents/) looks at which of the products we have opened actually do, and whether they ask before acting.

## Personal AI agent names move, and so does ownership

One entry on this list has already been acquired: Poke, whose maker [joined Cognition](https://cognition.com/blog/interaction) in July 2026. The company behind another, Manus, maker of Cue, agreed a merger with Meta that was then broken off, as [TechCrunch reported](https://techcrunch.com/2026/09/18/manus-seeks-4b-valuation-in-new-500m-fundraise-as-it-resumes-independent-ops) in September. And at least one shares its name with an unrelated product: OpenAI's Dot and the Dot app from Emotion Computer are both here, as separate entries.

That is ordinary for a category this young, and it is a practical problem rather than a trivial one. Search for one of these products and you can land on a different company with a similar name, a predecessor that was acquired, or a community project that reimplemented it. We have had to separate at least one pair by hand while assembling this page, and the notes on individual entries record which is which where it is not obvious.

It is also why the source column exists. A link you can open settles which product an entry means, in a way a name on its own does not.

## The personal AI agents we are not showing

A competitor publishes a longer list. We read it, and some of the names on it are not here; the note under the table gives the current count.

Those are names we could not trace to an official site or a first-hand report in the time we spent looking. The 9 October checks found mostly unrelated products sharing those names; none was traced to a personal agent. They are held back because listing a name nobody can open is the kind of filler that makes a list longer without making it more useful, and because the whole point of this page is that every line can be checked.

If one of them is yours, or you simply know where it lives, that is exactly the sort of correction worth sending, and it is the fastest way for an entry to appear here.

## What it takes to move up a state

The three states are not a queue we work through in order, and nothing is promised about how long a product stays in one.

Moving from listed to read from source needs published code and documentation. A runtime teardown needs access to a running product. Several of these are invite only or waitlisted, so the wait is not ours to shorten. Once inside, the work is watching what the product actually does rather than what its documentation says: where the model runs, what it can reach, what it keeps between sessions, and which of those we can establish with evidence rather than infer. Every teardown we publish ends with a list of what we failed to establish, and that list is usually the longest section.

Moving from examination to scored needs something different again. The exam has to be able to drive the product without a person in the loop, which not every one of these allows, and the tasks have to run in worlds that can be reset and replayed so two attempts are comparable. A product that can only be operated by hand can be examined but not scored, and that is a property of our exam rather than a verdict on the product.

Neither promotion is automatic and neither is a favour. A product sitting at listed for a long time usually means we could not get in, or could get in and could not establish anything we were willing to publish.

One consequence worth stating plainly: the four products with scores are not the four best, or the four biggest. They are the four that were reachable when the exam was built. Reading the scorecard as a ranking of the category would be reading it as something it cannot be, which is exactly why this page exists beside it.

## How this list of personal AI assistants gets used

It is not only a page. The same file drives what the [radar](https://personalagentbench.com/radar/) watches, so adding a product here puts it in front of the collector the next morning, and third-party findings about it become publishable rather than being rejected as off topic.

Before 2026-09-24 that collector watched four products and nothing else, which meant a published finding about any other personal agent was invisible to us twice over: never searched for, and refused at the gate if it somehow arrived. The list fixed both, which is a better reason for it to exist than being a page.

## FAQ

**How do you decide what counts as a personal AI agent?** Something aimed at an individual rather than a team, that acts on their behalf rather than only answering questions, and that reaches real accounts or services to do it. The directory includes consumer agents alongside related workplace products, frameworks and memory components. A listing is discovery evidence, not a claim that every entry is a shipping consumer assistant.

**Why is there no score next to most of these?** Because they have not been through the exam. Four have, and which of them has a rate depends on the batch: Town's September figure is withheld until it reconciles with the records we can check, and in the 11 October batch Pine was set aside because almost none of its submissions reached the exam. Putting a number next to the others would mean inventing one, and a benchmark that does that has stopped being a benchmark.

**Which personal AI agent is best?** We do not answer that, on this page or any other. The scorecard has measured completion rates for the products that have been through the exam, and the [teardowns](https://personalagentbench.com/agents/) explain how each one is built.

**Does listing a product mean you recommend it?** No. The list records that a product exists and how it is reached. Several entries are ones we know very little about, which is precisely what the listed state means.

**How often does this change?** The date at the top is when the list was last checked. New products appear in this category faster than they can be examined, so treat the list as current to that date rather than as complete.

Pine focuses on phone errands, Instinct works through messaging, Muse combines a cloud VM with apps, and Town offers scheduled routines and native Mac access. Their [teardowns](https://personalagentbench.com/agents/) explain what was observed; the wider list includes products covered only through published sources.

---

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.