Criteria for choosing the best personal AI agents
Choose an agent for one repeatable job, then check what happens at the point where it could send, spend or delete. This is a documentation comparison checked 2 October 2026, grouped by use case rather than performance rank. We make OperatorNest; its entry includes the same controls and buying limits as the others.
- Works while you’re away: a schedule, an event trigger and a local desktop session have different requirements. Ask whether the task runs with your devices off.
- Channels: distinguish a place you can message the agent from an account it can read. A Slack connector does not establish that you can delegate through Slack.
- Model choice: separate a user-facing picker, automatic routing and bring-your-own keys or subscriptions. Ask whether switching preserves history.
- Approvals: identify the exact send, purchase, deletion or publication that pauses. Check whether a recurring task inherits an earlier grant.
- Memory and export: ask what is stored, what you can inspect or correct, and what you can take to another service. Deleting an account and exporting memory solve different problems.
- Receipts and logs: look for sources checked, actions taken, approvals and failures. An enterprise audit log may have different access requirements from your task history.
- Availability and price: check your country, account type, device and required plan. Monthly USD figures below are published prices, with usage charges or separate subscriptions called out where documented.
A detail marked unverified means the cited material did not establish it. It does not mean the feature is absent. Capabilities below are vendor descriptions, not results from a shared hands-on test.
A shortlist by job
All facts and prices in this table were checked on 2 October 2026. The product sections link to sources and explain eligibility. On a phone, scroll sideways to see every criterion.
| Product and fit | Background work | Where you reach it | Model control | Approvals | Memory and record | Published price |
|---|---|---|---|---|---|---|
| Meta Muse: WhatsApp admin | Schedules and events | Apps, web, WhatsApp | Meta model; no BYO documented | Cards; adjustable caution | Editable memory; activity log; export unverified | Free for most needs; paid prices unverified |
| Poke: text delegation | Recipes; paid automations | Messages, WhatsApp, Telegram | User choice unverified | Universal gate unverified | Memory; export and action-log coverage unverified | Free; Pro $19/mo; Ultra $199/mo plus usage |
| Instinct: contextual follow-ups | Proactive follow-ups | Text and calls | Unverified | Universal gate not promised | Context collection; memory tools and receipts unverified | Unverified |
| ChatGPT agent / Work: project tasks | Schedules and events | Desktop, web, mobile | OpenAI picker; no external BYO documented | Check-ins and app permissions | Memory controls; enterprise logs | Plan dependent; Pro $100/$200/$500 monthly |
| Gemini Spark: Google tasks | Cloud schedules; local needs device | Gemini mobile, Mac, web | No external BYO documented | Web confirmation; full scope unverified | Connected context; memory export and receipts unverified | US Pro $19.99/mo; Ultra $99.99/mo |
| Grok Bot: persistent computer | Routines and events | Desktop, mobile; team Slack | Managed selection; no picker | Configurable review | Shared files; enterprise logs; export unverified | Bundled paid access; allowances vary |
| Lindy: inbox and meetings | Scheduled briefs | Slack, iMessage, connected inbox | Per-task choice; BYO unverified | External sends/writes wait | Workspace context; action logs; export unverified | From $29.99/user/mo |
| Manus: BYO-key projects | Schedules and events | Web, Telegram | Flex BYO keys and model controls | Send/publish confirmations can be skipped | Persistent context/files; memory export and receipt coverage unverified | Free; Pro from $20/mo plus Flex inference |
| Genspark: cross-channel Claw | Cloud schedules and monitoring | Workspace; Claw chat channels | Claw picker; BYO unverified | Claw external-action permission | Cross-channel memory; no chat-history export; receipt coverage unverified | Plus $24.99/mo; cloud Claw costs extra |
| Perplexity Computer: research | Recurring and conditional tasks | Web, desktop, mobile, email | Multi-model; BYO inference unverified | Check connector grants | Brain editing on eligible plans; export and receipt coverage unverified | Pro $20/mo; Max $200/mo |
| OperatorNest: independent recurring work | 24/7 schedules and watches | Chat apps, email, web | Any provider; own keys/subscriptions | Consequential actions wait by default | Read/edit/export/delete memory; action receipts | No public pricing; access by request |
Best for delegation from messages
Meta Muse
Best for: WhatsApp users who want personal admin without choosing a model.
Meta’s product guide describes scheduled and event-triggered work, readable and editable memory files, an activity log and approval cards. Its business announcement adds connected business accounts and says sends, publishing and spending need approval.
Falls short: no BYO model is documented, caution settings can change approval behavior, and a standalone memory export remains unverified. Availability is US and Canada. Price, checked 2 October 2026: Meta says free for most needs with paid subscriptions; current paid amounts remain unverified. Our Muse comparison.
Poke
Best for: people who want to delegate through Messages, WhatsApp or Telegram.
Poke describes connected-app tasks and Recipes. Its FAQ covers email/calendar connections, custom tools and website creation. The published Pro plan includes background automations.
Falls short: end-user model choice, portable memory export and a universal AI pre-action gate remain unverified. The FAQ’s human-assistance controls should not be treated as proof of the AI’s approval policy. Country eligibility and action-log coverage also remain unverified. Price, checked 2 October 2026: Free, Pro $19/month, Ultra $199/month with usage charges. Our Poke comparison.
Instinct
Best for: people considering contextual follow-ups and bookings through text and calls.
Instinct’s site describes proactive assistance and offers a direct text entry point. Its optional context sources include connected accounts and device information.
Falls short: its terms authorize connected-service actions without promising confirmation before every send. Disconnecting a source does not delete indexed copies; removal is a separate request. Memory inspection/export tools and task receipts remain unverified. Availability and price, checked 2 October 2026: users must be 18+; public prices, model choice and country eligibility remain unverified. Our Instinct comparison.
Best for work inside an existing platform
ChatGPT agent
Best for: ChatGPT users who want longer tasks grounded in project context.
OpenAI’s agent help directs users from the original agent mode to Work. Work supports scheduled and event-triggered tasks across eligible desktop, web and mobile surfaces, with an OpenAI model picker. The Work admin FAQ explains app permissions and enterprise observability.
Falls short: access depends on plan and workspace; no external-model BYO is documented. Memory controls support review and deletion, but cross-provider portability remains unverified. Price, checked 2 October 2026: Pro tiers are $100/$200/$500 per month; these are published Pro prices, not a claim that Pro is the minimum Work plan. Our ChatGPT comparison.
Gemini Spark
Best for: eligible Google users delegating tasks involving their Google apps.
Google’s Spark help distinguishes cloud work that continues with devices off from local sessions that need an awake computer. Spark supports schedules, connected Google context and web-action confirmation.
Falls short: no external-model BYO is documented; portable memory export and complete approval coverage remain unverified. Help requires a personal account, age 18+, eligible Pro/Ultra and Keep Activity, and excludes EEA, Nigeria, Switzerland and UK. Business eligibility is inconsistent across Google’s pages. US price, checked 2 October 2026: Pro $19.99/month; Ultra $99.99/month. Our Spark comparison.
Grok Bot
Best for: people who need a persistent computer and shared team bots.
The FAQ documents desktop/mobile access, routines and shared persistent files. Team Bots adds Slack access on Teams/Enterprise. Model selection is managed; there is no customer-facing picker.
Falls short: bots on one account share a computer, files and logins; separate role conversations are not security boundaries. Security documentation describes configurable review and enterprise logs. Action recording is opt-in, and standalone memory export is unverified. Price and access, checked 2 October 2026: bundled with paid individual Cursor plans and Teams, with eligible SuperGrok access; allowances vary and country eligibility remains unverified. Our Grok Bot comparison.
Best for connected work and research
Lindy
Best for: inbox follow-ups and meeting work with an external-action approval boundary.
Lindy’s pricing FAQ describes Slack/iMessage access, per-task model choice and scheduled reports. It says read-only work on approved sources proceeds automatically while external sends and writes wait for approval. Its security page says each action is logged.
Falls short: BYO inference and portable memory editing/export remain unverified. Work pauses when credits run out. Self-serve access is advertised; country eligibility remains unverified. Price, checked 2 October 2026: Plus $29.99/user/month, Pro $99.99, Max $199.99, with plan-specific credits. Our Lindy comparison.
Manus
Best for: multi-step projects where you want BYO API keys and model controls.
Manus Flex lets you choose provider, model and reasoning settings. Manus 2.0 adds event-triggered Automations and a persistent Cloud Computer; Telegram delegation is available across tiers.
Falls short: scheduled-task settings can skip send/publish confirmations. General memory export, complete receipt coverage and country eligibility remain unverified. Price, checked 2 October 2026: Free includes two scheduled automations; Pro starts at $20/month. Flex inference is billed by your provider; other resources consume Manus credits. Our Manus comparison.
Genspark
Best for: cross-channel work through Claw alongside a document workspace.
Genspark Claw help documents model selection without clearing history, cross-channel memory, WhatsApp/Slack/Teams/Telegram access, schedules and Heartbeat monitoring. It says external sends and posts require permission.
Falls short: local Claw needs the computer/app running; cloud Claw needs a separate computer subscription whose current price remains unverified. Chat-history export is explicitly unavailable; BYO inference, complete receipt coverage and country eligibility remain unverified. Price, checked 2 October 2026: Free; Plus $24.99/month; Pro $249.99/month, before the cloud-computer charge. Our Genspark comparison.
Perplexity Computer
Best for: research that should turn into files and recurring connected-app work.
Computer help describes multi-model research, background jobs and conditional triggers across web, desktop and mobile. Brain gives eligible Max/Enterprise Max users source-linked memory they can inspect, edit and delete.
Falls short: memory export and BYO inference remain unverified. Review connector grants before recurring work; a connector permission is not evidence that every consequential action pauses. Complete receipt coverage remains unverified. Price, checked 2 October 2026: Pro $20/month; Max $200/month. Paid-plan access is documented; countries remain unverified. Our Computer comparison.
Best for independent recurring work
OperatorNest
Best for: people who want 24/7 delegated work with their own model access and portable memory.
OperatorNest runs schedules and watches from chat apps, email or the web. You can choose a model from any provider and bring your own keys or subscriptions. Memory can be read, edited, exported or deleted; switching models keeps it. Consequential actions wait for approval by default, and receipts record actions and decisions. Specific tasks can have scoped pre-approvals.
Falls short: access is by request, and no public pricing lets you compare total cost before contacting us. If one task already fits an eligible product you pay for, keeping it there avoids another setup. Receipts still need review; they do not guarantee correct research. Price and availability, checked 2 October 2026: no public pricing; access by request. Confirm the channels and connected accounts your job needs during setup.
Example: one job to check before handing over more
For example, Sam wants a weekday inbox follow-up brief at 7:00. The instruction is: “Read the connected inbox, list replies I owe, draft three nudges, and hold every send for my approval. Show which messages you checked and any failures.”
Check a run while the laptop is off, reject one draft, and inspect what the agent records. Correct a remembered recipient preference, then ask how to export that correction. If a missing inbox connection produces an empty brief without reporting the failure, fix the task before making it recurring. That check answers more about this job than a polished booking demo.