Productivity

How to Build Custom Integrations on Zo

A common question we get from new users is "Does Zo integrate with X?"

The latest productivity tool. An internal CRM. A weather API. A smart home device. The list of services people want to wire into their Zo never ceases to amaze me, and it grows every week.

Most people use dozens of services that don't always talk to each other. The promise of an AI agent is that it bridges that gap. The constraint, on most platforms, is that the agent can only reach the services on a fixed list.

Zo is built differently. The answer to "does Zo integrate with X" is almost always yes, because the model isn't a connector menu. Each Zo Computer is a real Linux server with persistent storage, environment variables, network access, and an AI agent that can run any command on it. If a service exposes an API, an SDK, a webhook, a CLI, or even a web page, your Zo can reach it.

This post breaks down the three paths to building a custom integration, when to use each one, and how to take a working integration from "something Zo did once" to "something Zo does reliably every time."

The Three Paths to Any Integration on Zo

Integrating your favorite services with Zo typically falls into one of three categories, but they aren't mutually exclusive. Most useful workflows end up using more than one.

  • Path 1: Built-in apps and pre-configured tools. One-click OAuth for the apps most people use daily, plus 50+ tools your agent can call from the first message. Setup time is seconds.
  • Path 2: Conversation-driven custom integrations. For anything not in the built-in list, you hand your Zo the docs and the credentials and ask it to wire the integration up in plain English. Setup time is minutes.
  • Path 3: Skills. Reusable, packaged behaviors that load automatically when relevant. Skills turn a working integration from a one-off into a permanent capability. Setup time is one conversation, then forever.

The progression usually goes: try a built-in app first. If it doesn't exist, build the integration in conversation. If you'll use it more than once, package it as a Skill.

PropertyBuilt-in app/toolConversation-drivenSkill
Setup timeSecondsMinutesOne conversation, then permanent
Reusable across sessions
Auto-triggered
ShareableNativeYes (folder, GitHub, Skills hub)
Best forCommon services: Gmail, Calendar, Linear, Drive, Notion, StripeOne-off tasks, prototyping, exploring a new APIAnything you'll do more than twice

Path 1: Built-in Apps and Tools

The fastest integration is the one you don't have to build. From the first chat, your Zo arrives with two things already wired up.

Native one-click apps

Zo connects to the apps most people use every day through standard OAuth in Settings > Integrations. Today that includes Gmail, Google Calendar, Google Drive, Google Sheets, Google Tasks, Linear, Notion, Spotify, Dropbox, Airtable, OneDrive, Outlook, and X (Twitter), with more added regularly based on what users ask for. After you connect an account, your agent acts on it without you remembering the integration's name. "Find emails from my accountant about Q3," "block off Friday afternoon for deep work," "create a Linear issue for the bug I just hit," all just work.

50+ pre-configured tools

These ship with every Zo Computer:

  • Web and research: real-time web search, deep research, X/Twitter search, Google Maps, page reading, and a full browser with persistent logged-in sessions for any site you've signed into.
  • Media generation: image generation, image editing, video generation, diagram generation, audio and video transcription, image search.
  • Files and shell: read, write, and edit any file in your workspace; code search; parallel and sequential shell commands on a real Linux server you own.
  • Communication: send yourself SMS, email, Telegram, or Slack messages from any agent or scheduled task.
  • Hosting and publishing: ship a React page, an API endpoint, or a long-running service to your own zo.space and zocomputer.io URLs without leaving chat.
  • Automations and sub-agents: schedule recurring tasks, create dedicated sub-agents with their own personas, write rules that fire on context.
  • Commerce: create Stripe products, generate payment links, and accept real money through Stripe Connect.

We watch what users ask for and build the most-requested integrations directly into the platform, so the default keeps growing. If what you need is in the built-in list, you're done in 30 seconds. If it isn't, the next two paths still apply.

Path 2: Build a Custom Integration in a Conversation

You don't need to write code, set up a development environment, or know what an HTTP header is to build a working integration with a service that's brand new to your Zo. You need three things:

  • The service's API or SDK documentation. A URL is fine. For complex SDKs, you can paste in the README or upload a PDF of the technical reference.
  • An API key or auth credential. Stored as a secret in Settings > Advanced so it never appears in chat.
  • A clear description of what you want. Plain English. The agent reads the docs, writes the code, runs it on your machine, and reports back.

This works because Zo isn't pattern-matching against a library of pre-built integrations. Your agent reads the docs the same way a human developer would, then writes the code, then runs it on a real machine you own. That's the structural difference between Zo and traditional automation tools like Zapier: Zapier connects pre-built apps to each other, your Zo writes the integration on demand.

A 10-Minute Walkthrough: Wiring Up a New API

Suppose you want your Zo to check the weather every morning and text you if rain is in the forecast. The service is OpenWeatherMap, which isn't in the built-in apps. Here's the full path from "I had an idea" to "running on a schedule, forever."

Step 1: Store the credential

Sign up at OpenWeatherMap, copy your API key, and add it as a secret named OPENWEATHER_API_KEY in Settings > Advanced. Your Zo now has the credential without it ever appearing in chat.

Step 2: Build the integration in conversation

Send your agent a message like:

Prompt

The OpenWeatherMap API docs are at https://openweathermap.org/current. I want to check the current weather and a 24-hour forecast for Phoenix, AZ. My API key is in the OPENWEATHER_API_KEY environment variable. Read the docs and figure out how to call it.

Your Zo opens the docs, identifies the right endpoints, builds a working call, and shows you sample output. If the API has gotchas (geocoding, units, rate limits), it surfaces them. If anything looks wrong, you correct it the same way you'd correct a colleague: "The temperature looks like it's in Kelvin, fix it."

Step 3: Schedule it

Open Automations and create a new automation:

  • Name: "Morning Rain Check"
  • Schedule: Every day at 6:30 AM Phoenix time
  • Prompt: "Check the OpenWeatherMap forecast for Phoenix, AZ for the next 24 hours. If rain is expected, text me a one-sentence summary. If not, do nothing."

That's it. A working production integration. You didn't write any code or configure any servers. You read the OpenWeatherMap homepage, copied your API key, and had a conversation.

What About More Complex Services?

The same workflow handles genuinely complex integrations: Stripe, Twilio, internal company APIs, banking aggregators, webhook-driven services. The pattern is identical:

  • Find the docs.
  • Get your Zo the credentials safely (always via Settings > Advanced).
  • Describe the workflow in plain English.
  • Verify the result and iterate.

The only thing that scales with complexity is how much context you hand over. For a simple REST endpoint, a URL to the docs is enough. For a complex SDK, you might paste the full README, upload a PDF of the technical reference, or link to a tutorial. Your Zo uses it as a working spec.

A note on OAuth: services that require OAuth typically need you to register an app with the provider first and configure redirect URLs and scopes. That part lives outside Zo. Once you have credentials, the rest of the workflow is the same.

Path 3: Skills

The conversation-driven approach is fast, but every conversation starts fresh. If you build an OpenWeatherMap integration on Monday, your Zo doesn't automatically remember the exact endpoint structure or the unit conversions on Wednesday. You can rebuild it, but you're doing the work twice.

Skills solve this. A Skill is a small folder with instructions, scripts, and assets that your Zo loads on demand. When the right kind of conversation comes up, your Zo activates the Skill and uses what's inside.

  • Reusable. Build the integration once, use it forever, across all conversations.
  • Auto-triggered. A well-written Skill description means your Zo loads it when relevant, without you having to remember it exists.
  • Bundle scripts. If your integration needs a Python or TypeScript helper, the Skill carries it.
  • Shareable. A Skill is just a folder. Share it directly, publish it to a Git repo, or install one from the Zo Skills hub.
  • Composable. A Skill can call other Skills and combine with built-in tools.

The skill-creator Skill

Building a Skill from scratch is straightforward but tedious: there's a folder structure, a frontmatter spec, a description that triggers reliably, and a workflow for testing it across realistic prompts. You can do all of that yourself by reading the Agent Skills specification, or you can let your Zo do it.

The official Zo Skills hub publishes a Skill called skill-creator whose job is to help you build new Skills. It walks your Zo through capturing intent, drafting the SKILL.md file, writing test prompts, and iterating until the Skill works reliably. It's the Skill for making Skills.

To install it, open the Zo Skills hub in your Zo app and click Install on skill-creator.

Once installed, packaging an integration as a Skill looks like this:

Prompt

I built a working OpenWeatherMap integration in our last conversation. Use the skill-creator Skill to package it as a reusable Skill called weather. It should trigger any time I ask about the weather, forecasts, or rain.

Your Zo runs the skill-creator workflow, drafts the Skill, sets up test cases, and saves it to your Skills folder. The next time you mention weather, your Zo loads the Skill and the integration is back, identical to how you left it.

Skills aren't limited to API integrations. Browse the in-app Zo Skills hub for ideas on what other people have built and what you can build yourself.

Choosing the Right Path

The three paths map to three different operational realities, not three different skill levels. The right path depends on how often you'll run the workflow and how much you want to think about it later.

  • Use a built-in app or tool when the service is on the list. There is no reason to build something Zo already ships with.
  • Build it in a conversation when you're exploring a new service, doing a one-off task, or prototyping a workflow you might not need again.
  • Package it as a Skill the moment you realize you'll do it more than twice. The cost of making a Skill is one conversation. The cost of rebuilding the same integration ten times is ten conversations.

The shortest path to figuring out whether Zo integrates with the tool you care about is to ask your Zo. Paste a link to the docs, describe what you want, and watch what happens. For more on the infrastructure that keeps integrations running 24/7, see our guide to personal AI agents, or check pricing to pick a plan.

Build the integration you want

Drop a doc link, drop your API key, and tell your Zo what you want. The integration shows up in minutes.

Frequently Asked Questions

Does Zo integrate with my SaaS tool?
If your tool has an API, an SDK, or a webhook, yes. The built-in apps cover the most common services with one-click OAuth. For anything else, you can build a custom integration in a conversation by handing your Zo the docs and your API credentials.
Do I need to know how to code to build a custom integration?
No. You describe what you want in plain English, your Zo reads the docs, writes the code, runs it on your machine, and reports back. You correct mistakes the same way you'd correct a colleague: in conversation.
Where do I store API keys safely?
In Settings > Advanced, under the Secrets area. Add the key with a clear name like STRIPE_SECRET_KEY or OPENWEATHER_API_KEY. Your Zo reads it from the environment when it needs to. The value never has to appear in chat.
What's the difference between a custom integration and a Skill?
A custom integration is a workflow your Zo built in conversation. It works for that conversation, and you can rebuild it later, but it's not automatically loaded next time. A Skill is the same workflow packaged into a reusable folder with instructions, scripts, and a description. Skills load automatically when relevant and persist across all conversations.
Where do I get Skills?
The Zo Skills hub is built into your Zo app. Browse what's available, install Skills with one click, or publish your own. The skill-creator Skill helps you build new ones.
What if the integration I want involves multiple services?
Your Zo can chain calls across any number of services in a single workflow. Pull data from Stripe, enrich it with Clearbit, write the result to Notion, and text you a summary. Build the workflow once in conversation, package it as a Skill, run it on a schedule. Your Zo is the glue.
Is there a limit to how complex an integration can be?
The real limits are API access, auth restrictions, provider rate limits, and how clearly you can describe the workflow. Within those, your Zo can read long SDK documentation, manage webhooks, run long-lived background services, and orchestrate calls across any number of APIs.

Your Zo integrates with anything that has an API. Get started with Zo Computer.

More tutorials

Productivity

What Are Skills and How to Use Them

Agent Skills are an open standard for teaching an AI to do something the same way every time. A folder, an instructions file, and your AI picks it up automatically when it's needed.

Productivity

How to Make a Skill That Actually Works

A practical guide to writing, testing, and maintaining skills on Zo. When to make one, what makes the difference between a skill that fires reliably and one that quietly stops working, and how to keep skills useful as the work changes.

Productivity

Understanding the Differences Between Agent Skills and MCP Servers

Skills and MCP keep getting talked about like they're alternatives. They aren't. MCP is how an AI plugs into other apps. Skills are how it knows what to do. Most real setups use both.

Productivity

How to Automate Anything with Zo Agents

Set up AI agents on Zo that run on a schedule. Morning briefings, inbox summaries, price monitors, competitor tracking, and weekly reports, all on autopilot.

Productivity

Personal AI Agents: What They Are, How They Work, and Why 2026 Is the Year They Get Real

A technical breakdown of personal AI agent architecture in 2026: the observe-plan-act loop, persistent memory, tool integration via MCP, and why infrastructure, not intelligence, is the bottleneck.

Productivity

Agent Memory: A Plain-English Guide

A plain-English guide to AI agent memory in 2026. How vector search, knowledge graphs, and RAG actually work, and how the major memory providers differ in approach. Written by someone who uses this stuff every day.

How to Build Custom Integrations on Zo | Zo Computer