Hayz Labs · https://hayzlabs.com/reviews/top/automation-tools/
reviews · top 20 list · checked 18 Oct
Top 17 AI automation and agent tools
Tools that do multi-step work for you: workflows that run on their own, agents that browse and fill in forms, and tools that pull data from the web.
Best: ChatGPT (GPT-6) and Claude (tied, scoring 70). 2 of 17 scored.
Security notes for tools on this list
- Hermes Agent: warning. Three flaws this year: a malicious MCP catalog entry or git repo could run code (fixed in 0.19.0 and a later commit), and a medium flaw in the skills tool has no listed fix. Run it in a sandbox. Checked 18 Oct: source · more
- Langflow: warning. A critical flaw (CVE-2026-0768, 9.8) lets anyone who can reach the server run code without logging in. Attacks stealing API keys were reported from September 2026. Run 1.11.6 or later and never put it on the internet. Checked 18 Oct: source · more
- n8n. A user who can edit workflows could run system commands on the server, fixed in 2.31.5 and 2.32.1. Limit who can edit workflows. Checked 18 Oct: source
- OpenClaw. Two high-severity flaws this year, one exposing stored job secrets, fixed in 2026.8.1. It has many security advisories, so keep it updated and sandboxed. Checked 18 Oct: source · more
- Dify. An outsider could make the server fetch internal addresses (CVE-2026-105762), fixed in 1.13.0. Checked 18 Oct: source
- Perplexity Comet. Researchers showed in March 2026 that a calendar invite could silently hijack Comet's agent to read local files and steal saved passwords. Perplexity fixed it before disclosure; keep the agent out of logged-in accounts you cannot afford to lose. Checked 19 Oct: source
Scored ones first, highest score first; on a tie, the more widely used. Unscored ones follow, by adoption (the measure is named on each row). Open the Table to see the number that placed each one and how it was scored.
- 1ChatGPT (GPT-6)OpenAI
Good for agent work and coding on Plus with GPT-6 Astra. For writing, Gemini AI Pro and Claude Pro score higher.
Best for Long multi-step tasks across your apps and files on the $20 Plus planYou give up Free gets only limited Work on desktop; trains on your chats unless you opt out
70Checklist$20/mo - 2ClaudeAnthropic
Free runs Sonnet 5.5, near the top for coding and hard questions. Pay for Pro for agents and for writing with Opus 5.5.
Best for Agent work on your files and in Chrome for people paying for ProYou give up No agent tools on the free plan; usage limits are not published in numbers
70ChecklistFree - 3ManusManus
An agent that runs long web and file tasks for you; the free plan gives 300 credits a day.
Best for Handing long research and slide tasks to a cloud agentYou give up Not scored: only older versions tested; we could not read its training policy
Not scored yetFree - 4GensparkGenspark
An agent workspace that makes slides, sheets and docs from a prompt.
Best for Making slides, docs and sheets from one promptYou give up Credit use can add up fast; the free tier got a lifetime cap in Oct 2026
Not scored yetNo public price - 5Zapier (with Zapier Agents)Zapier
Easiest way to link thousands of apps without code; agents are priced separately
Best for Non-coders linking many everyday apps; the widest app coverage hereYou give up Free plan is 100 tasks a month; agent limits not shown; each tool call uses two tasks
Not scored yetFree - 6MakeMake
Visual automation builder for multi-step flows, with AI agents; a free plan to start
Best for Multi-step visual flows on a budget; cheaper per step than ZapierYou give up AI features may use extra credits; its AI agents are still in beta
Not scored yetFree - 7Hermes AgentNous Research
Free personal agent that remembers your work and runs from the terminal or chat apps; sandbox it before real use.
Best for Command-line users who want a free agent that learns their projectsYou give up Runs commands on your computer unless sandboxed; one flaw has no listed fix
Not scored yetFree - 8n8nn8n
Widely used tool for building AI automations and agents; free to self-host, cloud from 20 euros a month.
Best for Teams that want to self-host AI workflows and keep them on their serversYou give up Licence limits commercial use; sends usage data to n8n until you turn it off
Not scored yetFree - 9FirecrawlFirecrawl
Popular tool that turns web pages into clean text for AI apps; free to self-host, 1,000 free API pages a month.
Best for Developers whose AI apps or agents need to read web pagesYou give up Self-hosting lacks some cloud features; structured output costs 5 credits a page
Not scored yetFree - 10OpenClawOpenClaw Foundation
Popular free personal agent you run on your own computer; powerful but risky, so set up sandboxing before you use it.
Best for Command-line users who want an agent they can drive from chat appsYou give up Tools run on your computer unless you set up sandboxing; many security advisories
Not scored yetFree - 11DifyLangGenius
Visual builder for AI apps, agents and questions over documents; self-host free under its own license
Best for Building AI apps and document question answering in a visual editorYou give up Licence bans hosting it for many customers; paid cloud starts at $590 a year
Not scored yetFree - 12GumloopGumloop
Drag-and-drop AI workflows for ops, sales and marketing teams; no free plan
Best for Ops, sales and marketing teams building AI workflows without engineersYou give up No free plan, only a 14-day trial; Pro adds an 8% orchestration fee
Not scored yetFrom $37/mo - 13Browser UseBrowser Use
Python library that lets an AI model drive a real web browser to finish tasks on websites.
Best for Python developers who want a model to finish tasks on real websitesYou give up Its 97% score is self-reported; cloud bills add model cost on top
Not scored yetFree - 14LindyLindy
AI assistants for email, meetings and scheduling built without code
Best for People who want AI help with email, meetings and scheduling, no codeYou give up Free is $50 of credits for 7 days; Team costs $29.99 a month per user
Not scored yetFree - 15ActivepiecesActivepieces
MIT open-source alternative to Zapier and n8n with a free cloud plan
Best for Teams that want a fully open source alternative to Zapier or n8nYou give up Far fewer GitHub stars than n8n; enterprise parts use a commercial licence
Not scored yetFree - 16LangflowLangflow (IBM/DataStax)
Security warning: a critical flaw was attacked from Sep 2026; run 1.11.6 or later and never expose it
Best for Developers building agent and document flows that export to PythonYou give up Self-host only; a critical flaw has been attacked, so keep it updated and private
Not scored yetFree - 17Perplexity CometPerplexity
AI browser for Mac, Windows, iOS and Android that can do shopping, email and travel tasks
Best for People who want a free browser that can summarise tabs and do web tasksYou give up Hidden instructions can hijack its agent; keep it out of accounts you can't lose
Not scored yetNo public price
| # | Name | Score | What placed it | Price | Best for | You give up |
|---|---|---|---|---|---|---|
| 1 | ChatGPT (ChatGPT Work agent)OpenAI · Closed · our review | 70Our checklist | Ranked #2 of 52 in a public blind vote for agent tasks (arena.ai)What it rests onTheir figure: arena Agent #2 of 52 (GPT-6 Astra Max, +14.13%). reused review (agents): Plus with GPT-6 Astra, arena Agent #2 of 52 today -> job 2, model 2, value 1, privacy 1, limits 1 = 7 Report a mistake | Free (limited Work, desktop); Plus $20/mo | Long multi-step tasks across your apps and files on the $20 Plus plan | Free gets only limited Work on desktop; trains on your chats unless you opt out |
| 2 | Claude (Cowork and Claude in Chrome)Anthropic · Closed · our review | 70Our checklist | Ranked #1 of 52 in a public blind vote for agent tasks (arena.ai)What it rests onTheir figure: arena Agent #1 of 52 (Claude Opus 5.5 Max, +18.88%). reused review (agents): Pro on Opus 5.5, arena Agent #1 of 52 today -> job 2, model 2, value 1, privacy 1, limits 1 = 7 Report a mistake | Free; Pro $20/mo ($17/mo billed yearly) | Agent work on your files and in Chrome for people paying for Pro | No agent tools on the free plan; usage limits are not published in numbers |
| 3 | ManusManus · Closed | Not scored | manus.im ranks #2,433 of all websites by visits (Similarweb)NotesTheir figure: Similarweb global #2,433 (manus.im, Sep 2026). not scored: no independent test of Manus 2.0 in the last 90 days; Scale RLI lists only Manus 1.6 and older Report a mistake | Free (300 credits/day, 2.0 Lite); Pro from $20/mo | Handing long research and slide tasks to a cloud agent | Not scored: only older versions tested; we could not read its training policy |
| 4 | GensparkGenspark · Closed | Not scored | genspark.ai ranks #4,866 of all websites by visits (Similarweb)NotesTheir figure: Similarweb global #4,866 (genspark.ai, Sep 2026). not scored: no independent test of Genspark in the last 90 days Report a mistake | Not confirmed | Making slides, docs and sheets from one prompt | Credit use can add up fast; the free tier got a lifetime cap in Oct 2026 |
| 5 | Zapier (with Zapier Agents)Zapier · Closed | Not scored | zapier.com ranks #6,646 of all websites by visits (Similarweb)NotesTheir figure: Similarweb global #6,646 (zapier.com, Sep 2026). not scored: no independent test of Zapier Agents in the last 90 days Report a mistake | Free (100 tasks/mo); Professional from $19.99/mo; Agents Pro $33.33/mo billed yearly | Non-coders linking many everyday apps; the widest app coverage here | Free plan is 100 tasks a month; agent limits not shown; each tool call uses two tasks |
| 6 | MakeMake · Closed | Not scored | make.com ranks #8,083 of all websites by visits (Similarweb)NotesTheir figure: Similarweb global #8,083 (make.com, Sep 2026). not scored: no independent test of Make or its AI agents in the last 90 days Report a mistake | Free (1,000 credits/mo); paid plans | Multi-step visual flows on a budget; cheaper per step than Zapier | AI features may use extra credits; its AI agents are still in beta |
| 7 | Hermes AgentNous Research · Open source · our reviewSecurity warning: see above | Not scored | nousresearch.com ranks #10,276 of all websites by visits (Similarweb)NotesTheir figure: Similarweb global #10,276 (nousresearch.com, Aug 2026); GitHub 254.9k stars; Docker Hub 14.4M pulls. not scored: no independent test of Hermes Agent itself in the last 90 days; results depend on the model you pick Report a mistake | Free (open source, MIT); optional Nous Portal from $20/mo | Command-line users who want a free agent that learns their projects | Runs commands on your computer unless sandboxed; one flaw has no listed fix |
| 8 | n8nn8n · Open source · our reviewSecurity note: see above | Not scored | n8n.io ranks #12,682 of all websites by visits (Similarweb)NotesTheir figure: Similarweb global #12,682 (n8n.io, Sep 2026); GitHub 207.9k stars; Docker Hub 273.4M pulls. not scored: no independent test of n8n in the last 90 days Report a mistake | Free self-hosted; Cloud Starter EUR 20/mo, Pro EUR 50/mo billed yearly | Teams that want to self-host AI workflows and keep them on their servers | Licence limits commercial use; sends usage data to n8n until you turn it off |
| 9 | FirecrawlFirecrawl · Open source · our review | Not scored | firecrawl.dev ranks #25,991 of all websites by visits (Similarweb)NotesTheir figure: Similarweb global #25,991 (firecrawl.dev, Aug 2026); GitHub 192.7k stars. not scored: no independent test of web-extraction quality in the last 90 days Report a mistake | Free self-hosted; API Free 1,000 credits/mo, Hobby $19/mo | Developers whose AI apps or agents need to read web pages | Self-hosting lacks some cloud features; structured output costs 5 credits a page |
| 10 | OpenClawOpenClaw Foundation · Open source · our reviewSecurity note: see above | Not scored | openclaw.ai ranks #26,182 of all websites by visits (Similarweb)NotesTheir figure: Similarweb global #26,182 (openclaw.ai, Sep 2026); GitHub 392.4k stars. not scored: no independent test of OpenClaw itself in the last 90 days; results depend on the model you pick Report a mistake | Free (open source, MIT); you pay for the model | Command-line users who want an agent they can drive from chat apps | Tools run on your computer unless you set up sandboxing; many security advisories |
| 11 | DifyLangGenius · Open sourceSecurity note: see above | Not scored | dify.ai ranks #46,790 of all websites by visits (Similarweb)NotesTheir figure: Similarweb global #46,790 (dify.ai, Sep 2026); GitHub 158.8k stars; Docker Hub 24.9M (dify-api) pulls. not scored: no independent test of Dify in the last 90 days Report a mistake | Free self-hosted; Cloud Sandbox free; Professional $590/yr | Building AI apps and document question answering in a visual editor | Licence bans hosting it for many customers; paid cloud starts at $590 a year |
| 12 | GumloopGumloop · Closed | Not scored | gumloop.com ranks #96,789 of all websites by visits (Similarweb)NotesTheir figure: Similarweb global #96,789 (gumloop.com, Sep 2026). not scored: no independent test of Gumloop in the last 90 days (Cybernews review dated June 2026) Report a mistake | Pro from $37/mo (20k credits); 14-day trial | Ops, sales and marketing teams building AI workflows without engineers | No free plan, only a 14-day trial; Pro adds an 8% orchestration fee |
| 13 | Browser Use (Cloud)Browser Use · Open source | Not scored | browser-use.com ranks #112,396 of all websites by visits (Similarweb)NotesTheir figure: Similarweb global #112,396 (browser-use.com, Aug 2026); GitHub 118.4k stars. not scored: Online-Mind2Web 97% is self-reported; no independent test of Browser Use Cloud in the last 90 days Report a mistake | Free library; Cloud pay as you go, browsers $0.02/hr plus model cost | Python developers who want a model to finish tasks on real websites | Its 97% score is self-reported; cloud bills add model cost on top |
| 14 | LindyLindy · Closed | Not scored | lindy.ai ranks #154,618 of all websites by visits (Similarweb)NotesTheir figure: Similarweb global #154,618 (lindy.ai, Sep 2026). not scored: no independent test of Lindy in the last 90 days Report a mistake | Free ($50 credits for 7 days); Team $29.99/mo per user | People who want AI help with email, meetings and scheduling, no code | Free is $50 of credits for 7 days; Team costs $29.99 a month per user |
| 15 | ActivepiecesActivepieces · Open source | Not scored | activepieces.com ranks #157,900 of all websites by visits (Similarweb)NotesTheir figure: Similarweb global #157,900 (activepieces.com, Aug 2026); GitHub 25.1k stars; Docker Hub 2.35M pulls. not scored: no independent test of Activepieces in the last 90 days Report a mistake | Free (1,000 credits/mo); Plus $20/mo; self-host free | Teams that want a fully open source alternative to Zapier or n8n | Far fewer GitHub stars than n8n; enterprise parts use a commercial licence |
| 16 | LangflowLangflow (IBM/DataStax) · Open sourceSecurity warning: see above | Not scored | langflow.org ranks #227,663 of all websites by visits (Similarweb)NotesTheir figure: Similarweb global #227,663 (langflow.org, Aug 2026); GitHub 155.8k stars. not scored: no independent test of Langflow in the last 90 days Report a mistake | Free (open source, MIT) | Developers building agent and document flows that export to Python | Self-host only; a critical flaw has been attacked, so keep it updated and private |
| 17 | Perplexity CometPerplexity · ClosedSecurity note: see above | Not scored | No separate visit count (it runs inside perplexity.ai)NotesTheir figure: no separate traffic figure (runs inside perplexity.ai); placed last. not scored: no independent test of Comet's agent in the last 90 days (Cybernews test dated Oct 2025) Report a mistake | Not confirmed | People who want a free browser that can summarise tabs and do web tasks | Hidden instructions can hijack its agent; keep it out of accounts you can't lose |
Nothing matches all of that. , or try everything.
How we ranked this
Order. Scored ones first, highest score first; on a tie, the more widely used. Unscored ones follow, by adoption (the measure is named on each row).
How we score this list. Our checklist: independent evidence that it does the job (from the last 90 days), the model under it, value, privacy and reliability. The full rules are on How we score. Open "How it was scored" on any row to see the working.
What we did not do. We did not run any of these ourselves. Every number comes from the public source linked on its row, checked on 18 Oct 2026.
Spot a mistake? A wrong rank, price or link: tell us and we will check it against the source. Corrections are dated in News.
Reviews are drafted with help from AI models, including Anthropic's Claude. Model scores come only from outside leaderboards and independent evaluations, by a written rule. Claude (Cowork and Claude in Chrome) is made by Anthropic, whose Claude models help draft our reviews.