THE ITEM SHOPAI and automation tools, grouped by the job they do

ChatGPT dots

OpenAI's always-on agent inside ChatGPT: you give your dot a goal and it keeps working on it between conversations, on its own cloud computer, and comes back for approval when it needs you.

Breaks whenPro access excludes the EEA, Switzerland and the UK at launch, rollout is gradual, and you can only create a dot in the ChatGPT desktop app or desktop web, not on mobile.

Grok Bot

xAI's AI teammates: named Bots with their own jobs that run on a persistent cloud computer, sign into your apps and websites, and finish tasks while your laptop is closed.

Breaks whenAll your Bots share one cloud computer with the same files, browser sessions and logins, so xAI says not to treat separate Bots as security boundaries; one Bot can run only one computer-use task on screen at a time.

Gemini Spark

Google's 24/7 personal agent in the Gemini app: it runs tasks and schedules in Google's cloud across Gmail, Drive, Docs and Calendar, even when your phone and laptop are off.

Breaks whenPersonal Google Accounts only (no work or school accounts), 18 or over, Keep Activity must be on, and it is not available in the EEA, Nigeria, Switzerland or the UK.

Perplexity Computer (Personal Computer)

Perplexity's agent for multi-step work across the web, files and connected apps; the Personal Computer desktop app adds your local files and apps and, on a Mac mini, runs 24/7 in the background.

Breaks whenComputer runs on credits (100 credits = $1): light tasks use about 100 to 350 credits and the largest projects 2,400 to 9,800; when credits run out, active tasks pause.

Claude hand-off and scheduled tasks

Anthropic's Claude app (now merged with Cowork) takes a task or a schedule and works on it in the background, in the cloud or on your desktop, then messages you the finished document, deck or pull request.

Breaks whenIt is task and schedule based, not one named agent watching your accounts; tasks that need local files, Chrome or desktop apps need Claude Desktop open and the computer awake.

Manus Cue and Automations

Manus 2.0 runs work when a signal or schedule fires (Automations), and Cue is its separate app for a personal team of agents, each with its own email, phone number, wallet and computer.

Breaks whenCue needs an invitation code for now, and an agent's phone number is a separate subscription that counts toward a Cue weekly allowance.

Kimi Work Goal Mode and Kimi Claw

Moonshot's desktop agent Kimi Work can chase one goal for up to 24 hours in Goal Mode, and Kimi Claw is a hosted OpenClaw agent in the cloud you can chat with in groups.

Breaks whenKimi Work is in Beta and Moonshot says execution stability and delivery quality are still improving; the Mac app needs Apple silicon.

OpenClaw

Open source, self-hosted gateway that turns any model into an always-available assistant you message from WhatsApp, Telegram, Slack, iMessage, Teams and more, with scheduled jobs and tools.

Breaks whenYou run and secure it yourself: keep the Gateway loopback-only behind SSH or Tailscale, and any non-loopback bind needs gateway auth.

ChatGPT and the OpenAI API

OpenAI's assistant for everyday writing, research, analysis and coding, plus the API that lets you build the same GPT models into your own product.

Breaks whenOn Free, Go, Plus and Pro, chats are used to train OpenAI models unless you turn off "Improve the model for everyone"; rating a reply with thumbs up or down can still send that whole conversation to training.

Claude and the Anthropic API

Anthropic's assistant for writing, analysis and coding on web, desktop and mobile, plus the API for building Claude models into your own product.

Breaks whenUsage limits reset on a rolling five hour window, paid plans add weekly caps, and Anthropic says it may limit usage in other ways at its discretion, so there is no fixed message count.

Google Gemini and the Gemini API

Google's assistant in a standalone app, with paid plans that add usage and storage, plus the Gemini API for building Google's models into your own product.

Breaks whenFree API tier and AI Studio content is used to improve Google products and may be read by human reviewers, so Google says not to send sensitive or confidential data (EEA, UK and Swiss users get paid tier terms instead).

Grok (xAI)

xAI's assistant on web, iOS and Android with SuperGrok subscriptions, plus the xAI API for Grok text, image, video and voice models.

Breaks whenAPI rate limits depend on cumulative spend: a new team starts at Tier 0 and unlocks higher limits at $50, $250, $1,000 and $5,000; over the limit you get 429 errors.

DeepSeek

A Chinese AI lab's chat app and a low cost API for its DeepSeek V4 models, usable with OpenAI or Anthropic style clients.

Breaks whenDeepSeek's privacy policy says it collects, processes and stores personal data in the People's Republic of China and may use it to train its models; you have a right to opt out of training.

Kimi (Moonshot AI)

Moonshot AI's assistant for chat, agents, slides, docs and coding, plus the Kimi API for its K3 and K2 models.

Breaks whenA new API account must recharge at least $1 and starts at Tier 0: 1 concurrent request, 3 requests per minute and 1.5M tokens per day until $10 is recharged.

Mistral (Vibe, formerly Le Chat, and API)

A French AI lab's assistant and coding agent, Vibe (formerly Le Chat), plus Mistral Studio for its API and open-weight models.

Breaks whenVibe input and output are used to train Mistral's models by default unless you opt out; only Enterprise is opted out by default, and the pricing table lists Free, Pro and Team as opt-out.

OpenRouter

One API key and one OpenAI compatible endpoint for 500+ models from 80+ providers, with routing and fallback between them.

Breaks whenBuying credits costs a 5.5% fee on Standard and 8% on Business, on top of provider prices.

Perplexity

An answer engine that searches the web and replies with cited sources, plus APIs (Sonar and the Agent API) that bring search grounded answers into your own product.

Breaks whenSonar requests carry a per-request fee on top of tokens, $5 to $12 per 1,000 requests depending on model and search context size, so short queries cost more than the token price suggests.

Intercom (Fin AI Agent)

A customer support platform whose AI agent, Fin, answers customer questions across chat, email, voice and other channels before a human steps in.

Breaks whenCosts grow with success: every Fin outcome is billed from $0.99, and an assumed resolution counts when the customer simply leaves without asking for more help.

Zendesk (AI agents / Suite)

A full help desk and ticketing suite with AI agents that resolve requests across messaging, email and voice and hand the rest to human agents.

Breaks whenPlans include only a small allowance of automated resolutions (5 per agent per month on Team plans, 10 on Professional), and pay-as-you-go resolutions beyond that cost $2.00 each.

Help Scout (AI features)

An email-first shared inbox and help desk with built-in AI for drafting and summarizing replies and an AI Answers bot in its website widget.

Breaks whenAI Answers runs in the Beacon widget, so it is not a full AI agent across every channel.

Front

A shared team inbox for email and messaging channels with AI that triages, drafts and resolves conversations.

Breaks whenAutopilot only works with shared inboxes and is only optimized for English.

Gorgias

A help desk built for online stores that pulls store data into tickets and uses an AI agent to answer and sell to shoppers.

Breaks whenAn AI-resolved ticket is billed twice: once as a helpdesk ticket and again as an automated interaction ($0.90 annual or $1.00 monthly).

Superhuman Mail

A fast email app for individuals and teams that uses AI to sort, summarize and draft replies in your own Gmail or Outlook inbox.

Breaks whenIt only works with Gmail and Outlook accounts.

Clay

Build lead lists in a spreadsheet-style table and fill in missing details by pulling from many data providers and AI steps.

Breaks whenYou run out of actions mid-month: actions cannot be topped up separately, you must move to a higher plan (Clay docs).

Apollo.io

Search a built-in contact database, reveal emails and phone numbers, and run outreach from the same place.

Breaks whenYou need lots of phone numbers: an email or profile costs 1 credit but a mobile number adds 8 more, and unused credits reset each cycle with no rollover (Apollo docs and pricing FAQ).

Instantly

Send cold email campaigns from many inboxes at once, with warmup and a shared reply inbox.

Breaks whenYou expect one price to cover everything: outreach, leads and credits are separate plans, and the cheapest outreach plan caps you at 5,000 emails and 1,000 uploaded contacts a month (pricing page).

Smartlead

Send cold email at volume across many mailboxes, with warmup, sender rotation and client white-labeling for agencies.

Breaks whenYou outgrow the send cap: the $39 Base plan allows 2,000 contacts and 6,000 sends a month (pricing page).

NeverBounce

Check whether email addresses are real before you send to them, one at a time or in bulk lists.

Breaks whenYour list is full of catch-all domains: the API marks these as catchall, meaning unverifiable, so you still have to decide whether to send (API docs).

HubSpot (Sales Hub and free CRM)

Keep track of contacts, companies and deals in one CRM, with email tracking and meeting booking.

Breaks whenYou outgrow the free limits: 2 users, 1,000 contacts, 3 email templates, 10 custom properties and 2,000 email sends a month (HubSpot product catalog).

Lemlist

Run personalized cold outreach by email, and on the Multichannel plan by LinkedIn and other channels too.

Breaks whenYou want to send a lot from few inboxes: Lemlist's own guide caps campaigns at 25 to 30 a day per mailbox and 50 a day including warmup.

Common Room

Collect buying signals from website visits, community, product usage and social activity and tie them to specific people and accounts.

Breaks whenYou are a small team: the entry plan is $2,500 a month billed annually and there is no free plan (pricing page).

Zapier

Connects the apps you already use so that something happening in one app triggers actions in others, without writing code.

Breaks whenMulti-step workflows at volume get costly: every completed action is a task, and overage is billed at 2.5x the base rate on monthly plans (1.25x on annual).

Make

Builds visual, flowchart-style automations (called scenarios) that move and transform data between apps.

Breaks whenEvery module action counts as a credit, including reading, transforming, aggregating, and iterating rows, so loops over large lists burn credits fast.

n8n

Workflow automation you can run on n8n's cloud or host yourself, with code nodes for logic the built-in nodes do not cover.

Breaks whenSelf-hosting needs real setup: n8n's own docs mark technical knowledge as required and installation via npm, Docker, or a server, and you run the infrastructure yourself.

Retool

Builds internal tools such as admin panels and dashboards on top of your own databases and APIs.

Breaks whenCosts grow with headcount: Business is $50 per builder and $15 per internal user a month billed annually, and external users past 50 are billed per user.

Airtable

A spreadsheet-style database where teams store records and run simple automations on them.

Breaks whenRecord caps per base are firm: 1,000 on Free, 50,000 on Team, 125,000 on Business; you cannot add more records until you upgrade.

Activepieces

Open-source automation platform for building trigger-based flows and AI agents, on its cloud or on your own servers.

Breaks whenOn the Free plan, flows wait until the next day once the credits run out; paid plans bill overage at $0.007 per credit.

Lovable

Builds a working web app from a chat description, with hosting and a backend database included.

Breaks whenData access rules are wrong: CVE-2025-48757 found 170+ Lovable apps with missing or weak Supabase row level security, exposing user data and API keys to anyone with the public key.

Bolt.new

Turns a prompt into a website or app in the browser and publishes it live.

Breaks whenProjects grow: Bolt's docs say most token use comes from reading and syncing your project files, so larger projects cost more per message.

v0 (Vercel)

Generates React and Next.js interfaces and apps from prompts and deploys them on Vercel.

Breaks whenUsage is heavy: after the May 2025 move to credit pricing, a Vercel community thread said a simple 3 page form cost more than $1 and the $20 plan became impractical.

Replit (Agent)

An online workspace where an AI agent writes, tests and deploys apps for you.

Breaks whenThe agent touches production data: in July 2025 Replit's agent deleted SaaStr founder Jason Lemkin's production database during a code freeze, and he concluded there was no way to enforce a freeze (The Register).

Cursor

A code editor with AI built in, for developers writing and changing code in an existing project.

Breaks whenYou go past included usage: extra use is billed at API rates, and Cursor publicly apologized for unclear pricing in July 2025 and refunded surprise charges from June 16 to July 4.

Claude Code

An AI coding agent from Anthropic that reads your codebase, edits files, runs commands and opens pull requests.

Breaks whenYou hit plan limits: subscription usage has a session limit and a weekly limit shared with Claude chat, and when reached you wait, upgrade or pay for usage credits.

GitHub Copilot

AI code completion, chat and agents inside your editor and on GitHub.

Breaks whenThe task is big: a cloud agent session is capped at 59 minutes, works on one repo and one branch, and only on GitHub-hosted repos.

Appcues

Shows in-app guides, messages and surveys to your users so they find and adopt the features you ship.

Breaks whenEvery identified user who signs in counts toward your MAU total, whether or not they saw a flow, so cost follows your user count.

Userpilot

Combines product analytics with no-code in-app flows, surveys and a help center so you can see where users get stuck and guide them past it.

Breaks whenThe $299 Starter plan stops at 2,000 MAUs, 3 seats and 10 user segments; the next tier is from $849/mo.

PostHog

Records what users do in your product and bundles analytics, session replay, feature flags, experiments and surveys in one tool.

Breaks whenOn the free plan, events past the monthly allowance are permanently dropped.

Mixpanel

Analyzes product events so you can see which features people use, where they drop off, and who comes back.

Breaks whenOn the free plan, report access is blocked once you pass 1M monthly events until you upgrade or the cycle resets.

Customer.io

Sends automated messages by email, push, SMS, in-app and webhook based on what each customer does in your product.

Breaks whenBilling is on the maximum number of profiles in the account; reviewers note you pay for all contacts whether active or not.

Loops

Sends marketing, product lifecycle and transactional email for SaaS teams from one place.

Breaks whenIt is email only; there is no SMS, push or in-app channel.

Senja

Collects text and video testimonials from customers and shows them on your site as widgets and walls.

Breaks whenThe free plan stops at 15 testimonials and 1 project and shows Senja branding.

Rewardful

Runs affiliate and referral programs for SaaS by tracking who referred each paying customer and paying their commission.

Breaks whenPlans cap monthly affiliate-attributed revenue: $7,500 on Starter, $15,000 on Growth, and higher usage moves you up a tier.