Updated
Personal agents

Personal AI Agents

An agent that works for you rather than in your product — it reads your inbox, runs errands across your apps, does research and keeps going while you're away.

Also in Personal agents:Hosting for Personal Agents
For your framework or language:Next.js
Ask your AI about this, with this page as the source:ChatGPT ↗Claude ↗Perplexity ↗

Trends

Public counts each month: pull requests on GitHub, package downloads, stars, Product Hunt followers. They show direction, not market share.

ToolDownloadsStarsProduct Hunt
OpenClaw3.3M/wk★ 392.3k2.4k followers
Hermes Agent35.2k/wk★ 255.1k10 followers

The real choice

Run an open-source agent on a machine you control, or use one hosted by a model lab or a product company. Open-source agents such as OpenClaw and Hermes Agent cost nothing but model tokens and let you choose the models, but you keep a machine running, update it and decide what it may touch. Lab-hosted agents (Muse, Grok Bot, Perplexity Computer) give each user a managed computer and come with a subscription you may already pay for, but tie you to that lab's models and its rules on what the agent may do. Hosted products like Manus and Lindy sit in between: any model behind the scenes, integrations built for work, priced by credits.

What makers chose67 live products, each with a source. Not market share.

Pick by situation

Tap the ones that are you — the tools that fit light up below.
  • You want an assistant you message from WhatsApp or Telegram, on your own machine and your own model keysOpenClawHermes Agent
  • You want the agent to remember your projects and write its own reusable skills over timeHermes Agent
  • You want agents that each keep their own computer, browser logins and memory, and you're on a paid Cursor planGrok Bot
  • You're in the US or Canada and want a free agent that runs errands and shopping for youMuse
  • You want an agent other assistants can hand tasks to, through MCPPerplexity Computer
  • You want an agent for team work across many business apps without hosting anythingLindyManus
  • You want an agent on a schedule in Slack or Telegram without installing or hosting anythingBetterClaw

Compare the tools

Grouped by the side of the choice they answer, not ranked. Open a row for when to use it, the trade-off and what makers say.

ToolBest for
Open source, self-hosted2
OpenClawapp · open source · free tierA personal assistant you run on your own machine and talk to from WhatsApp, Telegram or another chat app, on the models you choose.64+28 open source3.3M/wk

Use it whenYou want full control over models, data and integrations, and don't mind keeping a machine online for it.

Trade-offRunning, updating and securing it is on you; it acts with your credentials, so what you connect is what it can reach.

Used by AgentCenter for OpenClaw, Blink.new, ClawMetry, ClawMote, ClawSecure and 59 more · in 28 open-source projects.

AgentPulse is built for the OpenClaw ecosystem. OpenClaw changed how we think about AI agents. We're building the management layer teams and agencies need to run them properly
Rectify, the makerApr 2026 ↗
Hermes Agentapp · open source · free tierA self-hosted agent from Nous Research with persistent memory, skills it writes for itself, scheduled tasks and subagents, reached from chat apps, a CLI or desktop apps.2+2 open source35.2k/wk

Use it whenYou want an agent that learns your projects over time and runs unattended jobs, under the MIT license.

Trade-offYou host it and pay for models separately; sandboxing is something you configure (local, Docker, SSH, Modal) rather than get by default.

Used by Memmy Agent, Clawdi · in 2 open-source projects.

The self-evolving Agent built by Nous Research; its practice in persistent memory and skill self-learning showed us that an Agent can "understand you better the more you use it".
Memmy Agent, the makerSep 2026 ↗
Hosted by a model lab2
Grok BotappNamed, persistent Bots that each keep memory and work on a cloud computer with a browser, files and terminal, and coordinate with each other.No evidence yet

Use it whenYou're on a paid Cursor plan or SuperGrok, or want agents you can also tag on X.

Trade-offAll of your Bots share one cloud computer and its logins, so access granted to one is available to all; only sold through Cursor or SuperGrok plans.

No maker's product we track shows it for this yet.

Museapp · free tierA consumer agent on its own secure VM that books, buys, fills forms and follows up, messaged from the Muse app or WhatsApp.No evidence yet

Use it whenYou're in the US or Canada and want an agent for personal and small-business errands, including connectors to Shopify, Stripe, Notion and Slack.

Trade-offOnly in the US and Canada so far, Meta's models only, and built for consumers rather than developers' workflows.

No maker's product we track shows it for this yet.

Hosted product2
ManusappA hosted general agent that carries out multi-step tasks — research, slides, websites, media — from a web, desktop or mobile app.5

Use it whenYou want results from one-off tasks without setting up integrations or a machine.

Trade-offClosed source and hosted only; the work runs in Manus's cloud with whatever accounts you connect to it.

Used by Driven, IonRouter, Lovie Formation, Plansera AI, Truely by Validia.

Used Manus to offload research and ops chores around the build and launch — pulling competitive notes together and turning loose ideas into actual task sequences instead of babysitting every step.
Plansera AI, the makerOct 2026 ↗
LindyappAn AI teammate connected to 1,500+ apps that works in Slack and Gmail, runs routines on a schedule and handles phone calls.No evidence yet

Use it whenThe work is a team's recurring business tasks — inbox, meetings, outreach, reporting.

Trade-offCredits run down with every action, from $29.99/month, and the agent lives in Lindy's platform rather than on your machine.

No maker's product we track shows it for this yet.

Before you choose

How to approach it

Start with one recurring chore that has a clear finish line — a weekly report, triaging an inbox — and give the agent only the accounts that chore needs. Turn on approval for anything that sends, buys or deletes, and read the agent's log for the first week before letting it run unattended. If you self-host, run it on a separate machine or VM, not the laptop that holds your keys.

Common mistakes
  • Giving the agent your main email, password manager and payment card on day one, so one bad instruction or injected web page can act on all of them.
  • Running a self-hosted agent on your everyday laptop with full disk access, where anything it reads in a message can make it touch your files.
  • Leaving scheduled tasks running with no spending cap on the model key, so a loop burns through tokens overnight.

Other options

Real choices most makers here won't need to weigh.

Decided alongside

What the 67 makers' products here chose for their other decisions.