If you didn’t already know, OpenClaw (formerly known as Clawdbot, then Moltbot—it’s had a bit of an identity crisis) is an open-source personal AI agent that runs directly on your own machine or a private server.
On ChatGPT or Claude, you type a question and get a response. But in OpenClaw, the AI agent actually does things on your behalf. It can clear your inbox, manage your calendar, browse the web, and a whole lot more.
It’s your personal AI concierge—it runs in the background of the apps you connect it to.
People are using it for all sorts of things:
- Ordering groceries
- Catching up on email
- Creating release notes
- Preparing daily briefings
- Automating recurring tasks
- …And dozens of repetitive tasks that never seem to leave your to-do list
OpenClaw handles all of this from whatever chat app you already use—WhatsApp, Telegram, Slack, iMessage, Discord, etc.—so you don't need to learn a new interface. You just message your agent like you’d message a colleague.
This article will help you understand what OpenClaw can actually do—specifically, which skills and APIs are worth adding to your agent and what you can build with each one.
{toc}
How to set up OpenClaw
Here’s what you need to set up OpenClaw:
- The machine you want to give OpenClaw full access to (Linux/macOS/WSL on Windows recommended)
- Node.js 22+ (installed automatically by the installer script)
- An internet connection (OpenClaw can’t run if you lose access to the internet or if your machine turns off)
- API key of your chosen LLM provider (OpenAI, Claude, Gemini, whatever you want to pick)
- A messaging channel (can be anything like Telegram, your WhatsApp number, Discord, or iMessage)
Once you have all that, here’s how you can start using OpenClaw:
- Open your Terminal or Command Prompt and run the one-liner installer from openclaw.ai. This downloads the CLI, installs it, and starts the onboarding wizard
- Your onboarding wizard will now ask you to choose your AI model connection(s) & enter its API key(s), configure gateway channels like Telegram or WhatsApp, and optionally enable skills/plugins to let the agent do tasks
This is the quickest way to get OpenClaw up and running. But there are several alternative methods for installing OpenClaw. For example, you can use a VPS for safety and start adding instructions (using SOUL.md) while setting up itself—more on that in the next section.
What are Skills and APIs in OpenClaw?
Before you can get the most out of OpenClaw, it helps to understand two terms that come up constantly: Skills and APIs.
- Skills are like a how-to manual for your AI agent. It teaches the agent how to use tools it already has access to, in the way you want it to. For example, the Gmail Skill is not just for giving OpenClaw access to your inbox—it’s also for teaching it which emails to flag, which to summarize in your daily briefing, and which ones you don’t care about.
ClawHub—OpenClaw's public marketplace—hosts a growing library of community-built skills.
- APIs are the external systems the agent talks to. If a Skill is the instruction manual, the API is the physical connection that lets OpenClaw actually talk to an external service.
An API (Application Programming Interface) is essentially a bridge between two apps. Every app you’d like to connect to OpenClaw (Google, Spotify, Notion, Trello, etc.) has an API. It’s only when OpenClaw connects to Gmail through its API that it gains the ability to read, sort, send, and manage your emails on your behalf.
Imagine OpenClaw is a phone operator. Skills are the operator’s hands (they do the pressing, typing, filing), and APIs are the phones/lines to other companies (the operator calls Gmail, Stripe, Google Calendar).
The elephant in the room: Security risks with OpenClaw
You’re giving OpenClaw access to everything on your machine, essentially. To perform actions, OpenClaw may require access to files, APIs, or services you connect to—which introduces real security considerations.
Plus, you might accidentally install a malware Skill that (which can be corrupted). There might also be malicious instructions in outside sources you give it access to (like emails and web pages).
Meta’s director of alignment at Meta Superintelligence Lab, Summer Yue, experienced this herself when her OpenClaw agent went rogue and started to delete emails from her inbox, despite her repeatedly instructing it not to.
So, how can you actually ensure your OpenClaw runs as securely as possible? Here are some practical tips to follow:
- Run it in an isolated environment. If possible, run OpenClaw on a dedicated machine (like a new Mac Mini) or inside a virtual machine (such as a VPS like Hostinger) rather than your main work laptop.
- Add approval gates for important actions. In your agent’s instruction file (called SOUL.md), add lines like: “Never send an email without showing me the draft first” or “Never delete files, archive instead.”
- Start with read-only access. Let your agent summarize emails before you let it send them. Let it read your calendar before it books meetings. Build trust incrementally.
- Never give it root access. Create a dedicated user account on your computer for OpenClaw with limited permissions.
- Vet every Skill before installing. Third-party “skills” (plugins that extend OpenClaw’s abilities) are a big source of risk because they may contain malicious code or instructions. Check who built it, how many stars it has, and read its ClawHub VirusTotal report. If a Skill for something simple (like checking the weather) asks for permission to run shell commands, that’s a red flag.
Now that you understand what OpenClaw is and how to use it safely, the next section will cover the most useful skills & APIs for OpenClaw agents and what you can build with each.
15 OpenClaw skills worth installing (and what to build with each)
Here are the Skills and APIs that OpenClaw users rely on most—along with real examples of what you can build with each one.
1: Google Workspace (Gmail, Calendar, Drive, Docs, Sheets)
Gog has long been the go-to Google Workspace skill on ClawHub, with over 14,000 downloads—giving your agent access to Gmail, Calendar, Drive, Docs, Sheets, and Contacts through a single connection.
More recently, Google itself published an official Google Workspace CLI on GitHub, containing guidelines for OpenClaw integration.
The key difference: gog is the simpler starting point for most users, while the CLI is worth switching to if you need deeper or more comprehensive Workspace access. It's labeled “not an officially supported Google product,” so use it with that in mind.
What you can build:
- Weekly status summaries pulled from your emails and Docs
- Meeting scheduler that checks your calendar and drafts invitations on your behalf
- Automatic email triage—the agent flags priority messages and files or archives everything else
- A newsletter filter to auto-archive marketing newsletters and only surface vendor/client mail
- A morning briefing that pulls your calendar events, unread emails, and flagged tasks before you pick up your phone
2: Documents and knowledge (Notion, Obsidian)
Skills: Obsidian, Notion, Apple Notes
If you use the above tools as a second brain or knowledge base, this skill connects your vault directly to your agent. OpenClaw can read your notes, add new ones, and link ideas across documents—all from a Telegram/WhatsApp/Discord message.
What you can build:
- Daily journal prompts that pull from your recent notes
- Research summaries that connect ideas across existing notes
- Project tracking that updates your vault as tasks are completed
- Automatic note capture from conversations—anything you tell your agent gets saved to the right folder
3: Web automation and scraping
Skill: Agent Browser
This Skill gives your agent the ability to navigate websites, fill out forms, extract data, and take screenshots—all autonomously.
What you can build:
- Automated flight check-ins
- Form submissions and repetitive data entry
- Screenshot monitoring of pages you want to track for changes
- Find leads by scanning a list of company pages, extracting contact details, and filling a spreadsheet
- Web scraping for competitive research—have your agent pull pricing or news from specific sites on a schedule
4: Code & files
Skills: GitHub
If you’re an engineer or work with software teams—even in a non-technical capacity—you can use the GitHub Skill to connect your agent to your repositories. It can monitor pull requests, summarize issues, and send you updates when something needs attention.
What you can build:
- Automatic notifications when CI/CD pipelines fail
- Create or triage issues directly from a Telegram message
- Daily digest of open PRs and issues across your projects
- Summarize PRs, suggest reviewers, and post short release notes
5: Team chat (Slack, Discord)
Use this Skill to let your agent send messages, react to posts, and monitor channels. It can also be used to post alerts, accept commands from the team, and escalate issues.
What you can build:
- Scheduled reminders to specific team members
- Automated status updates to your team’s daily standup channel
- Alerts when specific keywords appear in channels you care about
- Monitor payment failures or broken tests and post a summary to a channel
- Cross-channel summaries like, “what did the marketing channel discuss this week?”
6: Summarize
Skill: Summarize
With 10k+ downloads, this is one of the most universally useful skills. It lets your agent pull content from URLs, YouTube videos, podcasts, and local files and condense it into a clean summary.
What you can build:
- Podcast episode recaps delivered to your phone
- Daily news digest from RSS feeds or specific sites
- YouTube video summaries before you commit to watching
- Long document summaries for quick review before meetings
7: Project management
Skills: Linear, Trello, Asana, Todoist, ClickUp
If you’re juggling multiple projects across tools, this skill category turns your agent into a lightweight project manager. Instead of context-switching between Linear, Trello, and Notion to check what’s happening, you just ask—and your agent pulls the status, updates the board, or creates the ticket while you stay focused on the actual work.
What you can build:
- Send a weekly digest of open issues and blocked items to Slack every Monday
- Auto-create a ticket when your agent spots a bug in your logs or a complaint in your email
- Close tickets, update sprint status, or move cards across boards from a single chat message
- Pull all open items assigned to you before a standup and format them as a quick briefing
- Sync task status across tools—done in Linear, updated in Notion, notified in Slack, automatically
8: Health and fitness tracking
Skills: WHOOP, Apple Health, Fitbit
Most health apps give you data but don't do anything with it. Connecting WHOOP, Fitbit, or Apple Health to your OpenClaw agent changes that—it can pull your metrics, spot patterns across weeks of data, and actually adjust your schedule based on how recovered you are.
What you can build:
- Cross-reference health data with your work schedule to spot patterns like low recovery on late-meeting days
- Log meals and symptoms by chat, get a weekly pattern analysis in return
- Include HRV, sleep score, and recovery data in your morning briefing
- Auto-clear gym blocks when your recovery drops below a threshold
- Post a weekly health snapshot to a private Notion page
9: Payment and business systems
Skills: Stripe, PayPal + your CRM
Connecting your billing and CRM to your agent turns it into a genuine business operations tool. Instead of manually checking Stripe dashboards or updating customer records, your agent monitors revenue activity, triggers downstream workflows, and keeps your team informed—without anyone having to look.
What you can build:
- Trigger an onboarding sequence the moment Stripe logs a new purchase
- Post a daily revenue snapshot to your team’s Slack channel before standup
- Roll up monthly revenue, refunds, and failed payments into a shared Google Doc
- Flag cancellations or downgrades in your CRM with context from the customer’s history
- Get a Slack alert when a charge fails, with a support ticket opened in your CRM automatically
10: Observability and security skills
Skill: Security Auditor
Installing a skill is functionally the same as installing code on your machine. So, having runtime monitoring in place is the foundation that makes everything else safe to use.
What you can build:
- Run static security checks on any third-party skill before you install it
- Keep a full audit log of every action your agent takes, reviewable from Telegram or Slack
- Restrict which external servers your agent can contact so a rogue skill can’t exfiltrate data
- Lock your agent to a vetted set of approved skill bundles only, blocking everything else
- Schedule automated scans of all installed skills and flag anything that’s changed its permissions
11: Self-improving agent
Skills: Self-improving agent, Self-improving agent with self-reflection
This skill analyzes how you use your agent over time, identifies gaps in its capabilities, and builds new Skills—which you can review and activate. Once it’s running, your agent gets more useful every week.
What you can build:
- Automatically refine how your agent handles recurring tasks based on past outcomes
- Let your agent notice you always ask for the same Monday report and automate it without being told
- Have your agent write and install a new skill mid-conversation when it realizes it doesn't have what it needs
12: Email marketing (Kit, Klaviyo, Mailchimp)
Skills: Kit, Klaviyo, Mailchimp
If you run email campaigns, this skill connects your agent directly to your email marketing platform. Instead of logging into dashboards to check campaign performance or manually triggering sequences, your agent handles monitoring, reporting, and basic campaign management from your chat app.
What you can build:
- Pull a weekly list growth report and drop it into a shared Notion doc
- Get alerted when a campaign's open rate drops below your usual benchmark
- Trigger a specific sequence when a subscriber hits a tag or segment condition
- Ask your agent mid-conversation how last week’s broadcast performed without opening the platform
- Get a campaign performance summary—opens, clicks, unsubscribes—delivered to Slack after every send
13: Social media (LinkedIn, Reddit, X, Instagram)
Skills: LinkedIn, Reddit, X, Instagram Search
Use these Skills to get alerts about social media mentions, keep up with industry news & trends, and summarize all the info in your niche across platforms. There are also several skills to schedule social media content if you need it.
What you can build:
- Pull engagement stats from your last week of posts across platforms into a single summary
- Monitor social media for mentions of your brand, product, or competitors and get a daily digest
- Draft and schedule LinkedIn posts from a Telegram message, with your agent applying your preferred tone and format
- Get alerted when a specific keyword spikes on social media channels so you can respond to trending conversations in your niche
14: Smart home
Smart home skills connect your agent to the devices in your physical environment. Rather than using separate apps for lights, speakers, and sleep tracking, your agent becomes the single interface and can combine these integrations into routines that trigger automatically based on your schedule or health data.
What you can build:
- Control your entire home environment from a single Telegram message while traveling
- Dim the lights and start a focus playlist automatically when your calendar shows a deep work block
- Build a morning routine that turns on lights, starts music, and delivers your briefing at a set time
- Get a notification when your smart home detects something unusual, like a device left on overnight
15: News and media
Skill: News summary
If you need to stay on top of a specific topic, industry, or competitor without spending time scouring the internet, this Skill turns your agent into a passive monitoring layer. It watches sources you care about, filters out the noise, and delivers only what’s relevant—on your schedule, in your chat app.
What you can build:
- Track a specific topic across multiple sources and get a weekly digest with the most cited stories
- Monitor a competitor’s press coverage and get alerted the moment they announce something significant
- Get a morning briefing of headlines filtered to your industry or keywords, delivered before you open your phone
The skills above cover a lot of ground, but OpenClaw’s flexibility comes from its modular architecture—if a skill doesn’t exist, you can create one. And in most cases, you don’t need to write any code to do it.
How to build custom Skills in OpenClaw
There are various ways to build custom Skills in OpenClaw. Here are a few routes you can take:
Ask your agent
The most straightforward way to create a custom skill is to describe what you want in plain language and let your agent build it.
Tell it the task, the tool it needs to connect to, and any rules it should follow. OpenClaw will write the skill file, save it, and start using it in the same conversation.
One user described asking their agent to create a Todoist automation mid-conversation: “I wanted to automate some tasks from Todoist and Claw was able to create a skill for it on its own, all within a Telegram chat.” The agent handled the entire build without the user touching a config file.
Write a SKILL.md file
Every skill in OpenClaw is just a Markdown file with a clear description of what the skill does, what tools it needs access to, and any instructions the agent should follow when using it. Skills are Markdown files that contain instructions for agents to complete specific tasks and use specific tools—each skill is a directory containing a SKILL.md file with YAML frontmatter and instructions.
If you want more control over how a skill behaves, writing the SKILL.md yourself gives you that. You define the inputs, the outputs, the permissions it needs, and any guardrails you want in place. Once the file is saved to your skills folder, OpenClaw picks it up automatically in the next session.
A practical starting point is to describe your workflow to an LLM like Claude first, ask it to draft the SKILL.md, review it, and drop it into your workspace.
Use Zapier or n8n as a bridge when the API doesn’t exist
Not every tool you want to connect to has a public API that OpenClaw can reach directly. This is where Zapier and n8n become useful. Instead of building a direct integration, your agent triggers a webhook—a signal sent to Zapier or n8n—and the orchestration platform handles the connection to whatever app is on the other end.
The practical upside is that Zapier alone connects to over 6,000 apps. If you can describe the workflow (“when my agent sends this signal, do X in this tool”), you can almost certainly build it without waiting for a native skill to exist.
How to roll out OpenClaw (slowly)
The best way to start with OpenClaw is to install two or three skills that address your most repetitive tasks and build from there. Here’s a simple framework to get started:
- Start with one productivity integration: Email or calendar is usually enough to prove value immediately.
- Add one communication channel: Slack, Telegram, or Discord—so your agent can report back and receive commands.
- Introduce one automation layer: Something like web automation or project management—but only after you’re comfortable reviewing outputs.
- Avoid destructive permissions at first: Let your agent summarize, draft, and monitor before you allow it to delete, send, deploy, or modify files.
- Review behavior weekly: Check logs. Review actions. Adjust instructions in SOUL.md before expanding access.
As trust builds, use articles like this and the community on OpenClaw’s Discord to discover more skills and adapt them for your own workflow.
Just make sure you keep the security basics in mind: vet your skills, start with read-only access, and always review what your agent is about to do before you give it permission to act on anything high-stakes. OpenClaw is genuinely powerful—a little setup discipline upfront will make sure it stays working for you.
