AgentComparison
Comparison guide · vendor documentation, attributed

Personal assistant agents compared

Poke, Lucas, Muse, OpenAI dots, Instinct and Wabi all let you ask for personal admin in a chat. They differ in where you reach them, what they connect to, what they publish about price, and how clearly they describe approvals. This page puts those published facts next to each other. Everything comes from the vendor pages linked on each agent's record. We have not used or tested any of them, so nothing here says which one works best.

Checked 2 October 2026Vendor documentation, attributedNo hands-on tests

Side by side

AgentWhere you reach itPublished priceApproval statement
PokeApple Messages, Telegram, WhatsApp, RCS (Poke docs)Free $0; Pro $19/month; Ultra $199/month (monthly billing view)Connect supported integrations explicitly. Detailed write approval and retention controls need further verification.
LucasiMessage and WhatsApp (Lucas FAQ)Free tier; Pro $20/month or $192/year; Max $100/month or $960/year; 7-day reverse trialVendor says users approve connected accounts, external actions and data access
MuseMeta account; app or web (Meta help)Limited free use documented by Meta; paid price not establishedMeta recommends defining boundaries and reviewing actions
OpenAI dotsChatGPT-connected; phone messaging per OpenAI docsPrice not verifiedPlugin access is managed in ChatGPT
InstinctText or call (Instinct site)Price not verifiedPublic terms describe connected-service access and actions, and warn safeguards may not prevent unintended actions. Check your actual approval and account settings.
WabiiPhone app, group chats (App Store listing)Free download; plan prices not published on pages readWabi says you choose what it can access, it checks with you before important actions that need approval, and you can disconnect connections at any time

Unknown price means we found no vendor-published price on the pages we read. It does not mean free. Prices change, so confirm on the vendor page before paying. Muse's paid price is not established.

What each one documents doing

Poke

Poke documents an assistant in Apple Messages, Telegram, WhatsApp and RCS for email, scheduling, reminders and web search.

Documented tasks: Email reading, search and drafting; Calendar scheduling and availability; Reminders and web search; Apple Messages, Telegram, WhatsApp and RCS.

Best fit (documented): Messaging-based email, calendar and reminder work.

Watch for: Granting broad access before checking approval controls.

Vendor source · checked 2026-10-02

Lucas

Lucas describes a text-native iMessage assistant for email, calendar, documents, reminders and automations.

Documented tasks: iMessage interface; Email, calendar, documents and reminders; Requested briefings, nudges and follow-ups.

Best fit (documented): Personal administration through iMessage.

Watch for: Assuming a marketing summary proves every action asks for approval.

Vendor source · checked 2026-10-02

Muse

Meta documents Muse as a personal agent for actions, reminders, ongoing conversations and file work.

Documented tasks: Reminders and ongoing conversation context; PDFs, spreadsheets and other artifacts; App and web access.

Best fit (documented): Personal reminders, ongoing tasks and artifact work.

Watch for: Unmonitored actions where errors would be costly.

Vendor source · checked 2026-10-01

OpenAI dots

OpenAI documents dots using permitted plugins and shared ChatGPT memory, with data-access and action controls.

Documented tasks: Permitted plugins and shared ChatGPT memory; Private proactive research; Device access subject to device permission.

Best fit (documented): Personal connected-app work with shared ChatGPT context.

Watch for: A user needing per-memory inspection and deletion.

Vendor source · checked 2026-10-01

Instinct

Instinct describes a personal assistant reached by text or call, connected to applications and devices.

Documented tasks: Text or call interface; Connections to applications and devices.

Best fit (documented): A messaging-first assistant connected to personal apps and devices.

Watch for: Connecting sensitive services without checking controls and terms.

Vendor source · checked 2026-10-01

Wabi

Wabi 2.0 describes an AI messenger where you ask for plans, errands or small apps and it does the work, with friends or family in the same chat. Wabi says it can build a plan, tracker, map, list or game inside the chat and can raise things worth handling when you connect accounts.

Documented tasks: Handle errands and plans from a chat request, such as booking an appointment or planning dinner (as described); Group chats where everyone sees and moves the same plan; Create a plan, tracker, map, list or game inside the chat; Raise deadlines, bills or forgotten plans when you connect accounts.

Best fit (documented): People who want to coordinate plans and errands with others in one chat and are comfortable with an invite-only early product.

Watch for: Anyone who needs published pricing, approval rules or reliability data first, or who is not on iPhone.

Vendor source · checked 2026-10-02

Known gaps to read before you connect anything

How to choose without testing

  1. Start with the channel. If you only use iMessage, WhatsApp or Telegram, check each vendor's listed channels first. Poke lists Apple Messages, Telegram, WhatsApp and RCS. Lucas lists iMessage and WhatsApp. Wabi is an iPhone app per the US App Store listing.
  2. Check what you would connect. Email and calendar are the common case. Ask whether the vendor says it can send or only draft, and read the approval wording above.
  3. Look for a published price. Poke and Lucas publish tiers. Instinct, dots and Wabi did not on the pages we read.
  4. Read the data terms. Retention, training use and deletion differ. Disconnecting a service does not delete retained data for Instinct or dots, per their own policies.
  5. Start narrow. Connect one account, ask for a low-stakes task, and review what it did before granting more.

Pair this with the privacy and approval checklist and the free agents list. For one-to-one comparisons use the comparison tool, and see personal admin use cases. How we source and label claims is in the methodology.