# Best AI agents in 2026: 13 tools compared by job and price

> Best AI agents in 2026 by job: ChatGPT, Claude, Gemini, Manus, Genspark, Perplexity, Lindy, Zapier, Notion and coding agents, with prices checked Oct 18, 2026.

- Canonical: https://fluxgravity.io/guides/best-ai-agent
- Published: 2026-10-18
- Content type: guide

A shortlist of AI agents by job, with verified plan prices, what each vendor says it does and where each is limited, read from vendors' own pages on October 18, 2026. FluxGravity wrote this page and is listed last with a narrow fit.

![FluxGravity social card for a guide to the best AI agents in 2026, compared by job, price and limits.](https://fluxgravity.io/images/editorial/guides-best-ai-agent-editorial-map.svg)

*Original page-specific FluxGravity editorial map based on this guide's structure. It is not a product screenshot, account test, or performance result.*

## What are the best AI agents in 2026?

**Answer:** There is no single best AI agent, because the products do different jobs. For delegated work from a chat, the verified shortlist is ChatGPT (Work), Claude, Gemini (Spark), Manus, Genspark and Perplexity Computer, starting around $20 a month; for workflow automation it is Lindy, Zapier Agents and Notion; for coding it is Claude Code, Codex, GitHub Copilot, Cursor and Devin.

This page lists only agents whose official pricing and capabilities we could read on October 18, 2026. Each tool below has a plan price, what the vendor says it does, and its limits. We did not use any of these products, so the list is ordered by job, not by quality, and it makes no claim that one agent finishes work better than another.

FluxGravity, the publisher of this page, is listed last as a narrow fit for turning a brief into finished files such as documents, slide decks, spreadsheets, videos and research reports. It is not ranked above the others.

Prices and plans changed often in 2026, and several vendors state that features are in beta or rolling out. Treat the prices as a snapshot and open the vendor's own page before you pay.
- Delegate a deck, analysis or recurring update: ChatGPT Work ($20 Plus), Claude ($20 Pro), Gemini Spark ($19.99 Google AI Pro).
- Credit-based general agents with their own cloud computer: Manus (from $20), Genspark ($24.99), Perplexity Computer ($20 Pro).
- Workflow automation across apps: Lindy ($49.99 a user), Zapier Agents ($400 a year), Notion Agent ($20 a member).
- Coding agents: Claude Code ($20 Pro), Codex (ChatGPT Plus $20), GitHub Copilot cloud agent ($10 Pro), Cursor ($30 Premium for cloud agents), Devin ($20 Pro).

## How did we compare AI agents?

**Answer:** We compared each agent on what it does, what it costs, how usage is metered, whether it asks before acting, where it runs, and where it is limited, using only the vendor's own pricing, help and documentation pages read on October 18, 2026.

An AI agent here means a product that takes multi-step action for you, such as running a task in the background, using tools or connected apps, or changing files or code, instead of only answering a question. Perplexity puts the difference in one line: Perplexity answers your questions, and Computer does your work. OpenAI says to use Chat for an answer and ChatGPT Work to complete a task with a clear outcome.

We checked six things for every tool: the lowest paid plan and the free option, higher tiers and team seats, how usage is counted (credits, limits or activities), controls such as approvals, where the agent runs (cloud, desktop or your editor), and availability limits such as regions, betas and plan gates. We left a cell as not verified when the official page did not state it.

We did not run benchmark tasks, so nothing here measures accuracy, speed or output quality. Where a vendor claims a capability, the page says the vendor states it.
- What it does: the vendor's own description of the agent.
- Price: free option, lowest paid plan, higher tiers, team seat, in US dollars.
- Metering: credits, usage limits or activities, and what happens when they run out.
- Controls: approvals before actions, scheduling, cloud or local execution.
- Limits: regions, plan gates, betas, and facts that are not verified.

## ChatGPT: Work, dots and Codex

**Answer:** ChatGPT covers three agent styles under one subscription: Work for delegated tasks, dots for an always-on agent, and Codex for coding. Work and Codex are included in every ChatGPT plan from Free up, with Plus at $20 a month and Pro from $100; dots start on Pro.

OpenAI describes ChatGPT Work as a way to delegate real work. It can use your files, plugins and approved tools to create finished files and run workflows, you can follow progress and approve important actions, and on the desktop app it can also use local files, apps and the browser. Scheduled tasks repeat or monitor work, and cloud tasks do not need your computer awake.

Dots are a separate always-on agent that lives in the cloud with its own computer and browser and can keep working when your computer is off. You reach it in ChatGPT, Slack, Teams or by call. OpenAI says dots are rolling out gradually to Pro 100, Pro 200 and Pro 500 users over 18 outside the European Economic Area, United Kingdom and Switzerland, to Business Premium, and to Enterprise if an administrator enables them.

Work and Codex share one usage allowance. For Plus, OpenAI's estimates are local messages per five-hour period, such as 5 to 45 with GPT-6 Astra and 15 to 160 with GPT-6.1 Sol, and Pro plans currently have no five-hour limit. Plus and Pro users who reach a limit can buy credits instead of upgrading. Conversations with a dot do not count toward usage limits, but tasks it starts in Work or Codex do.
- Price: Free $0, Go $8, Plus $20, Pro from $100 ($100, $200 or $500), Business $20 per user billed annually ($25 monthly).
- Best for: delegating finished files, recurring updates and coding in one subscription.
- Limits: usage is shared with Codex and counted in estimates, not task numbers; dots are gated by plan, age and region.
- Not verified: how many Work tasks a plan completes.

## Claude: Cowork inside Claude, and Claude Code

**Answer:** Claude suits people who want chat, bigger delegated tasks and a coding agent in one plan: Pro is $20 a month ($17 billed annually), Max starts at $100, and Claude Code is included from Pro. Claude asks before taking an action by default.

On September 16, 2026, Anthropic said Claude Cowork and chat are merging into one Claude, so you can bring a quick question or hand over a report and Claude continues after you close your laptop. You can schedule a task to repeat, choose whether Claude checks in before acting, and create Docs, Slides and Designs in the conversation, which you can edit, present from Claude or download. The merge is rolling out to Pro and Max first, with Team and Free plans to follow.

Claude Code is Anthropic's agentic coding tool that reads your codebase, edits files, runs commands and connects to your development tools, in the terminal, an IDE, a desktop app or the browser. Free does not include Claude Code, and Pro and Max do.

Every plan has usage limits that reset on a rolling five-hour session window, and paid plans add weekly limits. Activity across Claude on web, desktop, mobile and Claude Code draws from one pool, so heavy coding can use the allowance you wanted for chat. Max gives 5x or 20x Pro's usage.
- Price: Free $0, Pro $20 ($17 a month billed annually), Max from $100, Team standard seat $20 billed annually ($25 monthly).
- Best for: one assistant for chat, scheduled tasks, documents and coding.
- Limits: usage pooled across chat and Claude Code; the merged agent is rolling out; the Free plan has no Claude Code.
- Not verified: task counts per plan.

## Gemini: Spark, Antigravity and Jules

**Answer:** Gemini Spark is Google's 24/7 personal agent, built into the Gemini app and integrated with Gmail, Docs and Slides. It needs Google AI Pro ($19.99 a month) or Ultra (from $99.99) in a supported country, and it asks before high-stakes actions.

Google describes Spark as a cloud-based agent that keeps working in the background when you close your laptop or lock your phone. You can set recurring tasks or triggers, teach it skills, and ask it to complete workflows such as turning meeting notes into a Google Doc and drafting the follow-up email. Google says it is designed to ask you first before high-stakes actions like spending money or sending emails.

Spark launched on May 19, 2026 in beta for U.S. Google AI Ultra subscribers. Google's August 13, 2026 post says it is available to Google AI Pro and Ultra subscribers in over 160 countries. Google's pages we read give no Spark usage numbers, so how many tasks a plan allows is not verified.

For coding, Google AI Pro lists entry rate limits for the agent model in Google Antigravity and higher limits for Jules, an asynchronous coding agent that Google calls beta. Antigravity is also built into the Gemini macOS app, where you approve a plan and the agent builds and tests code in the background.
- Price: Free $0, Google AI Plus $4.99, Google AI Pro $19.99, Google AI Ultra from $99.99 a month.
- Best for: people whose work runs through Google Workspace and who want a background agent.
- Limits: Spark needs AI Pro or Ultra and a supported country; Jules is in beta; Spark usage limits not verified.
- Business: Gemini Enterprise Business starts at $21 per seat a month.

## Manus

**Answer:** Manus is a general agent that works on its own virtual computer in the cloud and bills in credits. Free gives 1,000 registration credits and 300 daily credits, Pro starts at $20 a month for 4,000 credits, and Team starts at $20 per seat.

Manus describes itself in the App Store as a general agent that uses its own virtual computer to plan, research, write code and deliver a finished product, working asynchronously in the cloud so it keeps going after you close the app. Manus 2.0, announced September 28, 2026, adds Cloud Computer, Automations that start work when something happens in a connected service, a desktop Studio, a Video Editor and Game Dev add-ons, and Cue, an app for personal agents.

Pro includes Manus 2.0 Max, 20 concurrent tasks and 20 scheduled automations. Manus charges credits for the tokens, virtual machines and third-party APIs a task uses, so cost depends on each task. Manus's help center says it cannot yet predict a task's credit cost, and monthly credits do not roll over.

Manus's pricing page returned no readable plan text to us, so plan details come from its help center. Higher Pro tiers, annual dollar prices and Enterprise pricing are not verified. Cloud Computer ($30 or $50 a month) and website hosting are billed separately.
- Price: Free $0; Pro from $20 (4,000 credits); second tier from $40 (8,000 credits, 7-day free trial); Team from $20 per seat.
- Best for: multi-step research, reports, slides and websites you want delivered as files.
- Limits: credits per task are unpredictable; free daily credits work only with Manus 2.0 Lite.
- Not verified: higher Pro tier credits, annual dollar prices.

## Genspark

**Answer:** Genspark bundles a Super Agent with slides, sheets, docs, chat and code tools in one credit-based workspace. Free gives 100 credits a day, Plus is $24.99 a month for 10,000 credits, and Pro is $249.99 for 125,000.

Genspark describes Super Agent as an autonomous assistant that thinks, plans and acts to complete everyday tasks, and says you can start it, close your laptop and come back to finished work. Plus and Pro include full access to AI Slides, Genspark Code and more agents, a weekly Free Quota for Standard-mode tasks, 40+ models in AI Chat, and commercial use of AI-generated content.

Credits reset each billing period and do not carry over. Regenerations and failed agent retries cost the same as the original run, and AI Chat and AI Image have a separate 5-hour fair-use cap. Genspark's pricing page returned an access error to our fetch, so prices come from its help center, and the Free plan's lifetime credit allowance and Pro tier prices above $249.99 are not verified.
- Price: Free $0 (100 credits a day up to a lifetime allowance); Plus $24.99 ($19.99 a month billed annually); Pro $249.99 ($199.99 billed annually); Team $40 per seat.
- Best for: one workspace for research, slides, sheets, docs and chat with a credit budget.
- Limits: no credit rollover; 5-hour AI Chat cap; benefits dated through June 30, 2027.
- Not verified: credits per task, Free plan lifetime allowance.

## Perplexity Computer

**Answer:** Perplexity Computer is the agent layer on top of Perplexity's search: Pro ($20 a month) includes weekly Computer usage, Max ($200 a month) includes 10,000 credits a month, and further credits cost 100 to the dollar.

Perplexity describes Computer as an independent digital worker that completes tasks and workflows by synthesizing information across the web and your own work context, then crafting assets, creating functioning apps, managing emails and conducting research. It can monitor and act on a schedule, uses specialist sub-agents, and keeps persistent memory across sessions. Computer runs in a secure cloud sandbox, and on supported NVIDIA hardware Portable Computer runs on your own device without using credits.

Perplexity's typical credit ranges are 15 to 70 credits for a quick fact, 350 to 950 for a multi-step research report, 2,400 to 9,800 for a data dashboard and 15,000 or more for an entire app. Paid Pro plans include weekly Computer usage on the Light effort setting, rolling out gradually, and Max has a default monthly spending cap of $200. Free accounts do not have Computer.
- Price: Free $0; Pro $20 a month or $200 a year; Max $200 a month or $2,000 a year; Enterprise Pro $40 per seat.
- Best for: research-heavy work where cited search and delegated tasks sit in one product.
- Limits: credits per task vary; Pro's weekly Computer usage is rolling out gradually; Free has no Computer.
- Not verified: the size of the Pro weekly allowance.

## Lindy

**Answer:** Lindy is an agent teammate for inbox, calendar, meetings and Slack work. It starts with $50 in free credits that last 7 days, then costs $49.99 a month per active user for 5,000 pooled credits.

Lindy's pricing page describes a Slack-native teammate that works in threads and mentions, runs scheduled routines, keeps persistent workspace context, drafts replies in your voice, manages inboxes and prepares and follows up on meetings. It lists 40+ skills, 1,500+ integrations, MCP support, and model choice for each task. Anything with outside impact, such as sending an email, updating a ticket or publishing a doc, waits for your approval.

Credits measure the work Lindy does. Lindy lists everyday asks at 2 to 250 credits, deep work at 250 to 1,000 and big builds at 1,000 to 2,500. Monthly credits do not roll over, top-ups cost $10 per 1,000 credits, and if the pool runs low Lindy pauses credit-using actions instead of billing more. Every active user is billed, including people who @mention Lindy in Slack.

The page lists computer use as a feature but names Pro and Max plans that do not appear in its pricing cards, so which plan includes computer use is not verified.
- Price: Free $0 ($50 in credits for 7 days); Team from $49.99 a month for 5,000 credits; Enterprise custom.
- Best for: teams that want an assistant working in Slack, email and calendar with approvals.
- Limits: every active user takes a paid seat; credits do not roll over; free credits expire after 7 days.
- Not verified: which plan includes computer use.

## Zapier Agents

**Answer:** Zapier Agents builds specialized agents that act across 9,000+ apps you connect. Free allows 400 activities a month and Pro is $400 billed annually ($33.33 a month) for 1,500 activities, with Enterprise on custom pricing.

Zapier says you create specialized agents, with the help of Zapier Copilot, that connect to your business data and perform tasks across 9,000+ apps. Agents can only take actions in the apps you have connected and use the triggers and actions you have set up. Activities count actions an agent takes in behaviors or chat, web browsing and lookups from attached knowledge, and a Team account shares one pool of activities.

Zapier warns that AI does not give the same outcome every time because of non-determinism, and offers prompt tips for reliability. Enterprise adds a custom number of activities, agent sharing across the organization, audit logs and restricted-apps support. The monthly price for Agents Pro is not shown in the text we could read, so only the annual price is verified.
- Price: Free (400 activities a month); Pro $400 billed annually, $33.33 a month (1,500 activities); Enterprise custom.
- Best for: automating business workflows across many apps you already connect to Zapier.
- Limits: acts only in connected apps; results can vary run to run; activities are capped by plan.
- Not verified: Agents Pro monthly billing price.

## Notion Agent

**Answer:** Notion Agent completes multi-step tasks using context from Notion, your connected apps and the web, and it is included on the Business plan at $20 per member a month. Free and Plus get a limited trial, and Custom Agents are free to try, then $10 per 1,000 Notion credits.

Notion describes Notion Agent as work done for you: it completes complex, multi-step tasks using context from Notion, connected apps and the web. Business also lists AI Meeting Notes and Enterprise Search in beta. Custom Agents are agents that handle repetitive tasks autonomously and require Notion credits.

Notion says that yearly billing saves up to 20%, but the yearly per-member price is not shown in the text we could read. On non-Enterprise plans, Notion's AI providers keep a 30-day data retention, and Enterprise workspaces get zero data retention with those providers.
- Price: Free $0 (limited trial); Plus $10 per member a month (limited trial); Business $20 per member a month; Enterprise custom.
- Best for: teams whose documents, databases and meeting notes already live in Notion.
- Limits: Notion Agent is a Business feature; Custom Agents use credits at $10 per 1,000.
- Not verified: yearly per-member prices.

## GitHub Copilot cloud agent

**Answer:** Copilot's cloud agent works in the background on GitHub, which suits teams that already manage work in issues and pull requests. Copilot Pro is $10 a month with 1,500 AI credits, Pro+ is $39 with 7,000 and Max is $100 with 20,000, and Free does not include the cloud agent.

GitHub says the Copilot cloud agent can research a repository, plan changes and implement them in the background, editing files and running tests and linters in an ephemeral cloud development environment. You can start work from the agents panel, a conversation, or an issue or pull request, and you review changes and request refinements before creating a pull request.

Delegating tasks to third-party coding agents such as Claude and OpenAI Codex is in preview and is listed for Pro+ and Max only. Copilot usage is measured in AI credits, and Pro, Pro+ and Max can buy additional credits. Business is $19 and Enterprise is $39 per granted seat a month, with credits pooled across the organization.
- Price: Free $0 (limited, no cloud agent); Pro $10; Pro+ $39; Max $100; Business $19 and Enterprise $39 per seat.
- Best for: teams that delegate coding tasks from GitHub issues and pull requests.
- Limits: credits are used up at different rates by model; Free has no cloud agent; third-party agents need Pro+ or Max.
- Not verified: tasks per credit.

## Cursor

**Answer:** Cursor calls itself a coding agent for building ambitious software. Every plan includes the agent in some form, cloud agents and automations start on Premium at $30 a month, and third-party models in Cursor cost extra at API list price.

Cursor says agents turn ideas into code while you hand off tasks and focus on decisions, and it is available as a desktop app and a command-line interface. Cursor is now sold through The X Subscription, which bundles Cursor with Grok and X benefits, with Free, Plus ($8), Premium ($30), Super ($100) and Ultra ($200) plans.

Free lists limited agent requests. Cloud agents and automations, code hosting and AI code review start on Premium. Usage comes from a pool shared between Cursor and Grok that refreshes weekly, with Premium at 4x, Super at 15x and Ultra at 50x the Plus allowance. Teams are listed from $30 per user a month. Third-party models in Cursor, such as Claude or GPT, are billed at API list prices rather than included.
- Price: Free $0 (limited agent requests); Plus $8; Premium $30; Super $100; Ultra $200; Teams from $30 per user.
- Best for: developers who want an agent in a desktop editor or terminal with cloud agents on higher plans.
- Limits: shared weekly pool with Grok; third-party models billed separately; cloud agents need Premium or higher.
- Not verified: tasks per allowance.

## Devin

**Answer:** Devin positions itself as an autonomous AI software engineer. Free is $0 with a light quota, Pro is $20 a month with access to cloud agents (Devin Cloud), Max is $200 a month, and Teams is $80 a month plus $40 per full dev seat.

Devin's documentation says it is an autonomous AI software engineer that can write, run and test code, and suggests asking it to take on Linear or Jira tickets, new features and bug fixes. Its rule of thumb is that if you can do a task in three hours, Devin can most likely do it, which is a vendor claim we did not test. Devin runs in the cloud, in a desktop app and as a terminal tool that can hand off longer tasks to cloud Devin.

Each paid plan has a usage allowance that refreshes daily and weekly, and extra usage is bought at API pricing. Cost per message varies with the model, task size and reasoning. Max is described as significantly higher quotas than Pro.
- Price: Free $0 (light quota); Pro $20; Max $200; Teams $80 a month plus $40 per full dev seat; Enterprise custom.
- Best for: engineering teams that want to hand tickets to an agent in the cloud.
- Limits: allowances refresh daily and weekly; extra usage is billed at API pricing; vendor effectiveness claims are untested here.
- Not verified: tasks per allowance.

## FluxGravity (published this page): a narrow fit for finished files from a brief

**Answer:** FluxGravity wrote this page, and its fit is narrow: it is an AI workspace and agent that turns a brief into finished work such as documents, PDFs, slide decks, spreadsheets and graphs, whiteboard and explainer videos, motion graphics, research reports and social posts. Plans are Free ($0, 100 credits once), Spark ($5), Pro ($20) and Ultra ($50).

FluxGravity's Spark, Pro and Ultra plans give 650, 2,800 and 7,000 credits a month, and annual billing saves 20%: $4, $16 and $40 a month. Paid credits renew monthly and do not roll over. Free includes the documents and slides workflows and no spreadsheets and graphs, while the paid plans include all 39 live workflows, deep research reports from Spark up, and a social workspace on paid plans.

FluxGravity does not name the models behind its plans on its pricing page, and we have not compared its output with any other tool here. Coding is not among the workflows FluxGravity lists on its pricing page, so use a coding agent above for that job, and use FluxGravity only when the deliverable is a finished file or post.
- Price: Free $0 (100 credits once); Spark $5 ($4 billed annually); Pro $20 ($16); Ultra $50 ($40).
- Narrow fit: turning one brief into a document, deck, spreadsheet, video, research report or social post.
- Limits: credits renew monthly with no rollover; page, slide and job limits rise by plan.
- Not verified: output quality against other tools, because we have not tested it.

## Which AI agent should you pick for each job?

**Answer:** Pick by the job and the place your work already lives: ChatGPT or Claude for delegated documents and recurring tasks, Gemini Spark for Google Workspace, Manus, Genspark or Perplexity Computer for credit-based general agents, Lindy, Zapier or Notion for workflow automation, and a coding agent from the list for software work.

Start with the free or trial option of two tools for your job, run the same real task in both, and note how many credits or how much of your allowance it used. None of the vendors publishes credits-per-task numbers that would let us do this for you.

Check availability before you plan around a feature. ChatGPT dots, Gemini Spark and the merged Claude agent are rolling out by plan, age or country, and Manus, Genspark and Perplexity have credit rules that can change what a task costs.
- Delegate a deck, analysis or recurring update from a chat: ChatGPT Work (Plus $20) or Claude (Pro $20).
- An always-on agent that follows up between conversations: ChatGPT dots (Pro from $100) or Gemini Spark (Google AI Pro $19.99), where available.
- Work centered on Gmail, Docs and Slides: Gemini Spark.
- A credit-based general agent with its own cloud computer: Manus (Pro from $20), Genspark (Plus $24.99) or Perplexity Computer (Pro $20).
- Inbox, calendar and meetings in Slack with approvals: Lindy (from $49.99 a user).
- Automations across many connected apps: Zapier Agents (Pro $400 a year).
- Documents and databases in Notion: Notion Agent (Business $20 a member).
- Coding from GitHub issues and pull requests: GitHub Copilot cloud agent (Pro $10).
- Coding in a terminal, IDE or desktop editor: Claude Code (Pro $20), Codex (ChatGPT Plus $20), Cursor (Premium $30 for cloud agents) or Devin (Pro $20).
- A finished document, deck, spreadsheet, video, research report or social post from one brief: FluxGravity, which published this page.

## Which AI agents and tools did we leave out?

**Answer:** We left out any agent whose official pricing and capabilities we did not read on October 18, 2026, and we left out developer frameworks, which are libraries without consumer plans. Omission is not a judgment that a tool is weaker.

Other lists name AI coding assistants and agent tools such as CodeGPT, Postman AI, Snyk, Aikido and Replit, and agent frameworks such as LangGraph. We did not read their official plan pages in this pass, so we do not list them. Frameworks are code libraries that you assemble into your own agent, and they do not have one consumer plan price to compare.

We also did not cover Microsoft Copilot, Grok Bot (included on Grok Premium and above at $30 a month, covered in our Grok pricing guide), app-builder agents, or vertical agents for support, sales or security. Check the vendor's pricing page for those, and see our other guides for the plans we did verify.
- Left out: agents whose official pricing we did not read today.
- Left out: developer frameworks and libraries.
- Left out: vertical and enterprise-only agents without public prices.

## AI agent shortlist by job

Shortlist from vendor pages read on October 18, 2026, in US dollars. Order is by job, not a ranking. 'Not verified' means the official page did not state it. FluxGravity published this page.

| Job | Tool | Lowest paid plan | What the vendor documents | Where it runs |
| --- | --- | --- | --- | --- |
| Delegated work from a chat | ChatGPT (Work, dots) | Plus $20; Go $8 for limited use | Delegate decks, analyses and recurring updates; dots on Pro from $100 | Desktop and cloud; dots in the cloud |
| Delegated work from a chat | Claude | Pro $20 ($17 billed annually) | Merged Cowork and chat; scheduled tasks; asks before acting by default | Cloud, after you close the laptop |
| Delegated work from a chat | Gemini (Spark) | Google AI Pro $19.99 | 24/7 personal agent in Gmail, Docs and Slides; asks before high-stakes actions | Cloud |
| General credit-based agent | Manus | Pro from $20 | Own virtual computer; Manus 2.0, Cloud Computer, Automations | Cloud |
| General credit-based agent | Genspark | Plus $24.99 | Super Agent with slides, sheets, docs, chat and code tools | Cloud |
| General credit-based agent | Perplexity Computer | Pro $20 | Autonomous worker on top of cited search; scheduled tasks | Cloud sandbox; Portable Computer on your hardware |
| Workflow automation | Lindy | Team from $49.99 per user | Slack-native teammate with approvals and 1,500+ integrations | Cloud |
| Workflow automation | Zapier Agents | Pro $400 billed annually | Agents acting across 9,000+ connected apps | Cloud |
| Workflow automation | Notion Agent | Business $20 per member | Multi-step tasks using Notion, connected apps and the web | Inside Notion |
| Coding | Claude Code | Claude Pro $20 | Reads your codebase, edits files, runs commands | Terminal, IDE, desktop, browser |
| Coding | ChatGPT Codex | Plus $20 | Coding tasks locally and in the cloud, shared with Work usage | Desktop, CLI, IDE, cloud |
| Coding | GitHub Copilot cloud agent | Pro $10 | Works in the background on issues and pull requests | Ephemeral cloud environment |
| Coding | Cursor | Premium $30 for cloud agents | Coding agent in desktop and CLI; cloud agents | Desktop, CLI, cloud |
| Coding | Devin | Pro $20 | Autonomous AI software engineer with cloud agents | Cloud, desktop, terminal |
| Finished files from a brief | FluxGravity (publisher of this page) | Spark $5 | Brief to documents, decks, spreadsheets, videos, research reports and posts | Not verified |

Related references:
- [OpenAI pricing](https://learn.chatgpt.com/docs/pricing)
- [Claude pricing](https://claude.com/pricing)
- [Gemini plans](https://gemini.google/us/subscriptions/)

## AI agent prices side by side

US list prices read from each vendor's own pages on October 18, 2026, in US dollars before tax. 'Not verified' or 'not listed' means the official page we could read did not state it. FluxGravity published this page and is listed last; the order is not a ranking. Prices are plain facts, not a claim about cost per output.

| Tool | Free option | Lowest paid plan | Higher tiers | Team or business seat | How usage is metered |
| --- | --- | --- | --- | --- | --- |
| ChatGPT (Work, dots, Codex) | $0, limited | Go $8; Plus $20 | Pro from $100 ($100, $200 or $500) | Business $20 billed annually, $25 monthly | Shared Work and Codex limits, credits |
| Claude | $0, no Claude Code | Pro $20 ($17 billed annually) | Max from $100 | Team standard seat $20 billed annually, $25 monthly | Five-hour and weekly limits, usage credits |
| Gemini (Spark) | $0, no Spark | AI Plus $4.99; AI Pro $19.99 | AI Ultra from $99.99 | Gemini Enterprise Business from $21 a seat | Multiples of Free limits |
| Manus | $0: 1,000 credits once, 300 a day | Pro from $20 (4,000 credits) | From $40 (8,000 credits); higher not verified | Team from $20 a seat | Credits per task |
| Genspark | $0: 100 credits a day, lifetime cap | Plus $24.99 (10,000 credits) | Pro $249.99 (125,000 credits) | Team $40 a seat | Credits plus weekly Free Quota |
| Perplexity Computer | $0, no Computer | Pro $20 | Max $200 | Enterprise Pro $40 a seat | Weekly Computer usage plus credits |
| Lindy | $50 credits for 7 days | Team from $49.99 per user (5,000 credits) | More credits up to $9,999.50 a month | Every active user billed | Credits, top-ups $10 per 1,000 |
| Zapier Agents | Free: 400 activities a month | Pro $400 billed annually ($33.33 a month) | Enterprise custom | Team shares an activity pool | Activities |
| Notion Agent | $0, limited trial | Business $20 per member | Enterprise custom | Per member | Plan limits; Custom Agents $10 per 1,000 credits |
| GitHub Copilot | $0, no cloud agent | Pro $10 (1,500 AI credits) | Pro+ $39 (7,000); Max $100 (20,000) | Business $19; Enterprise $39 per seat | AI credits |
| Cursor | $0, limited agent requests | Plus $8; Premium $30 for cloud agents | Super $100; Ultra $200 | Teams from $30 per user | Weekly pool shared with Grok |
| Devin | $0, light quota | Pro $20 | Max $200 | Teams $80 plus $40 per full dev seat | Daily and weekly quotas, extra at API pricing |
| FluxGravity (publisher of this page) | $0 with 100 credits once | Spark $5 ($4 billed annually) | Pro $20 ($16); Ultra $50 ($40) | Not listed | Monthly credits: 650, 2,800, 7,000 |

Related references:
- [OpenAI pricing](https://learn.chatgpt.com/docs/pricing)
- [FluxGravity pricing](https://fluxgravity.io/pricing)

## What each vendor documents about controls and availability

From the vendor pages read on October 18, 2026. 'Not verified' means the official pages we read did not state it. Vendor claims are not tested here.

| Tool | Approval before actions | Scheduling or recurring tasks | Availability notes |
| --- | --- | --- | --- |
| ChatGPT Work | You approve important actions | Scheduled tasks | Included from Free; usage shared with Codex |
| ChatGPT dots | Asks for decisions that need you; admins control Enterprise access | Recurring tasks | Pro 100, 200, 500 over 18 outside EEA, UK, Switzerland; Business Premium; Enterprise |
| Claude | Asks before taking an action by default | Scheduled tasks | Merge rolling out to Pro and Max first |
| Gemini Spark | Asks first before high-stakes actions | Recurring tasks and triggers | Google AI Pro and Ultra in supported countries |
| Manus | Not verified | Automations and scheduled tasks | Pricing page unreadable; higher tiers not verified |
| Genspark | Not verified | Not verified | Pricing page unreadable; benefits dated June 30, 2027 |
| Perplexity Computer | Not verified | Scheduled tasks | Pro weekly Computer usage rolling out gradually |
| Lindy | Outside-impact actions wait for approval | Scheduled routines | Computer use plan not verified |
| Zapier Agents | Not verified | Behaviors and triggers | Acts only in connected apps |
| Notion Agent | Not verified | Custom Agents run autonomously | Notion Agent is a Business feature |
| GitHub Copilot cloud agent | You review changes before a pull request | Not verified | Not on Free |
| Cursor | Not verified | Automations on Premium and above | Sold through The X Subscription |
| Devin | Not verified | Scheduled sessions listed in docs | Allowances refresh daily and weekly |

## Frequently asked questions

### What are the best AI coding agents right now?

Desk research cannot rank them, but five have verified plans: Claude Code (included from Claude Pro at $20 a month), Codex (included in ChatGPT Plus at $20), GitHub Copilot's cloud agent (Pro $10, not on Free), Cursor (cloud agents from Premium at $30) and Devin (Pro $20, with cloud agents). Pick by where you code, then compare how much of each allowance a real task uses.

### Is ChatGPT still the best AI?

No vendor page can settle that, and we did not test any model. ChatGPT is one of several strong options: Work and Codex are included from Free, Plus is $20 and Pro starts at $100. Claude Pro is $20 and Google AI Pro is $19.99, and each covers different agent features.

### Who are the big 4 AI companies?

No official list defines the big four, and we did not verify market share. The model makers whose agent products we verified for this page include OpenAI (ChatGPT), Anthropic (Claude), Google (Gemini) and xAI, whose Grok Bot is covered in our Grok pricing guide. Treat the phrase as a media shorthand, not a ranking.

### What are the best AI tools?

It depends on the job. For delegated work, see ChatGPT Work, Claude, Gemini Spark, Manus, Genspark and Perplexity Computer. For automation, see Lindy, Zapier Agents and Notion. For coding, see Claude Code, Codex, GitHub Copilot, Cursor and Devin. For a brief turned into files, FluxGravity, which published this page, is a narrow fit.

### What is the difference between an AI agent and a chatbot?

Vendors draw the line at action. Perplexity says it answers your questions while Computer does your work, and OpenAI says to use Chat for an answer and ChatGPT Work to complete a task with a clear outcome. An agent runs steps, uses tools or connected apps and can continue after you close the window.

### Which AI agents have a free plan?

ChatGPT ($0, limited), Claude ($0, without Claude Code), Gemini ($0, without Spark), Manus (1,000 credits once and 300 a day), Genspark (100 credits a day up to a lifetime cap), Zapier Agents (400 activities a month), GitHub Copilot ($0, without the cloud agent), Cursor and Devin (light free quotas), Lindy ($50 in credits for 7 days) and FluxGravity (100 credits once). Notion offers a limited trial of its agent.

### Which AI agent is best for business teams?

Team prices differ in structure: ChatGPT Business is $20 per user billed annually, Claude Team standard seats are $20 billed annually, Notion Business is $20 per member, GitHub Copilot Business is $19 per seat, Genspark Team is $40 per seat, Manus Team starts at $20 per seat and Lindy bills every active user from $49.99. We did not test any, so match the one that fits your tools.

### Are AI agents safe to give access to my accounts?

Check each vendor's approval controls. Claude asks before taking an action by default, Gemini Spark asks first before high-stakes actions like spending money or sending emails, Lindy holds actions with outside impact for approval, and ChatGPT Work lets you approve important actions. Review important results, since vendors state that agents can make mistakes.

### How much do AI agents cost per month?

Entry paid plans in this comparison run from $5 (FluxGravity Spark) and $8 (ChatGPT Go) to about $20 to $25 for the standard tiers of ChatGPT, Claude, Gemini, Manus, Genspark and Perplexity. Heavy-use tiers run from $100 (ChatGPT Pro, Claude Max, GitHub Copilot Max) and $99.99 (Google AI Ultra) to $200 (Perplexity Max, Devin Max) and $249.99 (Genspark Pro). Lindy starts at $49.99 per user, and credit-based agents cost more if tasks are large.

### Do AI agents have hidden costs?

Yes, several. Manus and Perplexity charge credits per task and neither can give a fixed cost in advance, Lindy bills every active user and top-ups cost $10 per 1,000 credits, GitHub Copilot and Devin bill extra usage beyond the allowance, and Cursor bills third-party models at API list price. Read each vendor's limits before you commit.

### Which tools did you leave out?

We left out agents whose official pricing we did not read on October 18, 2026, including Microsoft Copilot, app-builder agents such as Replit and Lovable, CodeGPT, Postman AI, Snyk and Aikido, and agent frameworks such as LangGraph. Omission is not a judgment that they are weaker.

### How was this guide researched?

FluxGravity wrote it from each vendor's own pricing pages, help centers, documentation and release notes read on October 18, 2026, plus FluxGravity's own pricing page. We did not use or benchmark any agent, so it makes no claim about output quality. Where a page was unreadable or conflicted with another, the guide says not verified.

## Sources
- [OpenAI: Get started with ChatGPT Work](https://learn.chatgpt.com/docs/get-started-with-work)
- [OpenAI: ChatGPT and Codex pricing](https://learn.chatgpt.com/docs/pricing)
- [Anthropic: Claude plans and pricing](https://claude.com/pricing)
- [Google: Gemini subscriptions (United States)](https://gemini.google/us/subscriptions/)
- [Manus Help Center: What is the current membership pricing for Manus?](https://help.manus.im/en/articles/11711111-what-is-the-current-membership-pricing-for-manus)
- [Genspark Help Center: Membership Plans (updated Oct 16, 2026)](https://www.genspark.ai/helpcenter/membership-plans)
- [Perplexity Help Center: Using the Connector for Slack](https://www.perplexity.ai/help-center/en/articles/12167980-using-the-connector-for-slack)
- [Lindy pricing](https://www.lindy.ai/pricing)
- [Zapier pricing (Agents plans)](https://zapier.com/pricing)
- [Notion pricing](https://www.notion.com/pricing)
- [GitHub: Copilot plans and pricing](https://github.com/features/copilot/plans)
- [Devin pricing](https://devin.ai/pricing)
- [Cursor pricing (The X Subscription plan table)](https://cursor.com/pricing)
- [OpenAI: Meet dots](https://learn.chatgpt.com/docs/dots)
- [Anthropic: Claude Cowork and chat are now one Claude (September 16, 2026)](https://claude.com/blog/cowork-is-now-claude)
- [Perplexity Help Center: What is Computer?](https://www.perplexity.ai/help-center/en/articles/13837784-what-is-computer)
- [Perplexity Help Center: How credits work on Perplexity](https://www.perplexity.ai/help-center/en/articles/13838041-how-credits-work-on-perplexity)
- [Anthropic: Claude Code overview](https://code.claude.com/docs/en/overview)
- [Google blog: The Gemini app becomes more agentic (May 19, 2026)](https://blog.google/innovation-and-ai/products/gemini-app/next-evolution-gemini-app/)
- [Google: Gemini Apps release notes](https://gemini.google/release-notes/)
- [Google blog: Gemini 3.7 Flash (August 13, 2026)](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-3-7-flash/)
- [Apple App Store: Manus listing](https://apps.apple.com/us/app/manus-ai-agent-automation/id6740909540)
- [Manus blog: Introducing Manus 2.0 (September 28, 2026)](https://manus.im/blog/introducing-manus-2-0)
- [Manus Help Center: Discounts and special offers](https://help.manus.im/en/articles/11813668-are-there-any-discounts-or-special-offers)
- [Manus Help Center: Credit consumption rules](https://help.manus.im/en/articles/11711097-what-are-the-rules-for-credits-consumption-and-how-can-i-obtain-them)
- [Manus Help Center: Checking credit cost before a task](https://help.manus.im/en/articles/13185575-is-there-a-way-to-check-how-many-credits-a-task-will-cost-before-i-begin)
- [Manus Help Center: Cloud Computer plans and billing](https://help.manus.im/en/articles/15392078-understanding-cloud-computer-plans-and-billing)
- [Genspark Help Center: Super Agent](https://www.genspark.ai/helpcenter/super-agent)
- [Genspark Help Center: AI Chat](https://www.genspark.ai/helpcenter/ai-chat)
- [Genspark Help Center: Credits Guide (updated Oct 14, 2026)](https://www.genspark.ai/helpcenter/credits-guide)
- [Zapier Agents product page](https://zapier.com/agents)
- [GitHub Docs: Copilot on GitHub.com (cloud agent)](https://docs.github.com/en/copilot/concepts/copilot-surfaces/copilot-on-github)
- [GitHub Docs: Plans for GitHub Copilot](https://docs.github.com/en/copilot/get-started/plans)
- [Cursor home page](https://cursor.com/)
- [xAI docs: What is The X Subscription?](https://docs.x.ai/x-subscription/overview)
- [Devin docs: Introducing Devin](https://docs.devin.ai/get-started/devin-intro)
- [FluxGravity pricing](https://fluxgravity.io/pricing)
- [xAI docs: The X Subscription pricing](https://docs.x.ai/x-subscription/pricing)

## Related pages
- https://fluxgravity.io/pricing
- https://fluxgravity.io/guides/manus-pricing
- https://fluxgravity.io/guides/genspark-pricing
- https://fluxgravity.io/guides/perplexity-pricing
- https://fluxgravity.io/guides/grok-pricing
- https://fluxgravity.io/autonomous-ai-agent
- https://fluxgravity.io/ai-agent-for-work
- https://fluxgravity.io/ai-research-agent

## Try FluxGravity

Give FluxGravity one goal and receive finished work.

https://app.fluxgravity.io/