What if you could describe what you need done — in plain English — and an AI agent would just go and do it?
Not generate a draft for you to finish. Not suggest some steps for you to follow. Actually do the work.
That’s exactly what Errand AI does, and today I’m excited to share it with the world.
The Problem We All Share
We spend an extraordinary amount of time on tasks that are important but tedious — researching competitors, drafting and sending emails, gathering data from multiple websites, managing files across cloud services, updating GitHub issues, posting to social media, and coordinating information across dozens of tools and platforms.
AI chatbots can help with some of this. They can write a draft or summarise some text. But the moment you need something that involves multiple steps, interacts with real services, or requires actual execution — you’re back to doing it yourself.
What Errand AI Does Differently
Errand AI is a personal task automation platform built from the ground up for the age of AI agents. You describe a task in natural language, and an AI agent picks it up and works through it — step by step, tool by tool — until it’s done.
Think of it like sending someone on an errand. You say what you need. They go and do it.
Here’s what makes it powerful:
🤖 Real Execution, Not Just Suggestions Errand agents don’t just tell you how to do something — they do it. They browse websites with a full headless browser, send and manage emails, interact with GitHub repositories, search the web, manage files on Google Drive and OneDrive, and much more.
🔗 Built on the Model Context Protocol (MCP) Errand is designed around MCP — the open standard for connecting AI to tools and services. This means it’s endlessly extensible. Connect any MCP server and your agents instantly gain new capabilities. It also exposes its own MCP server, so your IDE, coding assistant, or other AI tools can delegate tasks to Errand programmatically.
🔄 Multi-Step Workflows Unlike simple prompt-response interactions, Errand handles complex, multi-step workflows. Need to research a topic across several websites, synthesise the findings, draft an email with the summary, and send it to your team? That’s one task.
🧠 Persistent AI Memory Errand agents remember context across tasks. They learn your preferences, recall past decisions, and get better at serving you over time. It’s not just automation — it’s automation that knows you.
📋 Multiple Interfaces Create tasks however works best for you:
- Web UI with a visual task board and real-time streaming of the agent’s work
- Voice transcription — speak your task and it’s converted to text automatically
- Slack integration — fire off tasks without leaving your workspace
- MCP endpoint — let your IDE or other AI agents create tasks programmatically
- Agent delegation — agents can spawn sub-tasks that run in parallel
⏰ Scheduling Schedule tasks to run later or on a repeating basis. Set up a daily research brief, a weekly report, or a recurring check on something that matters to you.
Your Data Stays Yours
This is the part I’m most proud of. Errand AI runs on your own hardware.
- Download the macOS Desktop app and you’re running in minutes
- Deploy with Docker Compose on any machine
- Scale with Helm charts on Kubernetes for team deployments
Container images are available for both x86_64 and ARM64 — Errand runs on everything from a Raspberry Pi to a production cluster.
Your tasks, credentials, and data never leave your machine unless you explicitly configure an integration that requires it. No data is stored on external servers. You’re in complete control.
An optional Errand Cloud subscription provides secure remote access from anywhere, but even then, your data passes through without being stored.
Isolated & Secure by Design
Every task runs in its own ephemeral container — a fresh sandbox created for the task and destroyed when it’s done. There’s no shared state between tasks, no persistent sessions, and no way for one task to interfere with another. You control exactly which tools and integrations each task can access.
Who Is This For?
Errand AI is for anyone who wants to reclaim the hours lost to repetitive digital work:
- Developers who want to delegate research, code reviews, issue management, and documentation tasks to an AI that can actually interact with their tools
- Knowledge workers who spend too much time gathering information, drafting communications, and coordinating across platforms
- Teams who want a shared task automation layer that respects data privacy and runs on their own infrastructure
- AI enthusiasts who want a self-hosted, extensible platform for experimenting with agent-based automation
Get Started
Errand AI is available now. Visit https://errand.sh to learn more, read the documentation, and get started in minutes.
Explore the project on GitHub: https://github.com/errand-ai
The future of personal productivity isn’t about better to-do lists or smarter notifications. It’s about having an AI that can actually go and run your errands.
Welcome to Errand AI. Tell it what you need. It’ll take it from here.
#AI #Automation #Productivity #AIAgents #MCP #ModelContextProtocol #SelfHosted #OpenSource #DevTools #FutureOfWork #TaskAutomation #ErrandAI