AI Marketing OSAI Marketing OS
OpenWorker logo
Platform
4.4Editor Rating

OpenWorker

Open-source desktop coworker that runs marketing, sales and ops jobs inside Slack, email and your CRM

Pricing

Free

Platforms

macOS, Slack, Web...

Developer

OpenWorker (MIT, open source)

Rating

4.4 / 5.0

Last Updated

September 27, 2026

Overview

OpenWorker is a local-first AI coworker rather than another cloud chatbot.

You install a desktop app, connect the accounts where work actually happens, and ask for an outcome - the Monday spend and attribution report, account research and call briefs, a reviewed pull request - instead of chatting for an answer.

Read on openworker.com on 27 September 2026, the roles are named explicitly as coworkers: a Marketing coworker covering "spend, attribution and the Monday report - without five dashboards", a Sales coworker that researches accounts, preps calls and drafts follow-ups, an Executive Assistant that owns the calendar and triages the inbox, and an Ops On-call coworker that triages alerts and drafts incident notes.

Security coworkers ship first.

Anything consequential - sending email, posting to Slack, opening a PR, running a command - arrives as a check-in you approve in the app or from Slack, and you decide how independently each coworker is allowed to act.\n\nThe commercial model is the reason a data-conscious marketing team would consider it.

Read on openworker.com on 27 September 2026, the app is open source and free, with no paid tier and no subscription: you pay your model provider directly through your own API key.

Supported providers listed on the site are OpenAI, Anthropic, Google, Z.AI GLM, DeepSeek, Kimi, MiniMax, Qwen, xAI, Mistral, Together AI, Fireworks and Ollama for fully local models, and you can switch model per task, even mid-conversation.

Credentials live in a local secret store on the device; an optional free OpenWorker account exists only because some vendors require the app to broker the OAuth handshake, and the site states tokens are handed straight to your machine rather than stored in the cloud.

Connectors cover Slack, Gmail, Outlook, Google Calendar, Notion, HubSpot, Attio, Google Drive, Dropbox, Box, GitHub, Jira, Linear and Asana, with triggers that start work from a Slack mention, an incoming email or a schedule - which is how a recurring report gets finished before standup instead of during it.\n\nThe project is real software rather than a landing page: the repository at github.com/andrewyng/openworker showed 18,295 stars, 2,585 forks and an MIT licence with a last push on 24 September 2026 when we checked it through the GitHub API on 27 September 2026, and topai.tools lists it among the fastest-growing AI Agents in its directory.

The trade-offs are worth stating plainly.

Security is the most mature coworker family, so the marketing and sales roles are newer and thinner by comparison.

There is no managed cloud edition, no team admin console and no built-in cost reporting - every user connects their own tools and pays their own provider bill, which is a policy decision your finance team has to make rather than a feature it gets.

The desktop app is macOS only today with Windows announced as coming soon, so Windows-based teams should wait.

Use Cases

Producing the weekly marketing report - spend, attribution and channel performance - from HubSpot, Google Drive and Notion on a schedule before the Monday standup

Turn a Slack mention into triaged work: read the runbook, gather the numbers and hold an in-thread reply for approval

Researching target accounts and drafting call briefs plus follow-up emails before an outbound push

Triaging an overflowing inbox and calendar, drafting replies and holding anything sensitive for human review

Answering campaign questions across CRM, docs and analytics data without opening five separate dashboards

Who Is This For

Marketing ops leads who want reporting, research and inbox work delegated without buying another seat-based SaaS

Privacy-sensitive teams that need agents to run on local models with data staying on the machine

Sales teams that still research every account manually before a call

Technical founders and growth engineers who prefer forking an MIT repository over a per-seat subscription

Mac-based revenue and marketing teams - and Windows users willing to wait for the announced build

Key Features

Coworkers Per Job

Role-based coworkers arrive with tools and working style preset - Marketing, Sales, Executive Assistant and Ops On-call - and you can describe any other job in plain language to build a custom one.

Bring Your Own Model

Run OpenAI, Anthropic or Google models on your own keys, open-weight options such as Kimi, GLM, DeepSeek or Qwen, or fully local models through Ollama - switching model per task.

Approval Check-Ins

Sending an email, posting to Slack, running a command or opening a pull request lands as a reviewable action you approve in the app or from Slack before it happens.

Triggers, Not Just Prompts

Work starts from a Slack mention, an incoming email or a schedule, so recurring jobs such as a weekly marketing report or an inbox triage run without anyone opening the app.

Local-First Connections

One-click connectors for Slack, Gmail, Outlook, Google Calendar, Notion, HubSpot, Attio, Google Drive, Dropbox, Box, GitHub, Jira, Linear and Asana, with tokens kept in a local secret store.

Pros & Cons

Pros

  • Free and MIT-licensed - no subscription, you pay only your model provider
  • Works inside Slack, email, calendar and HubSpot instead of adding another dashboard
  • Model-agnostic, including fully local models through Ollama for privacy-sensitive teams
  • Every consequential action is approval-gated, with credentials held on the device
  • Real open-source traction: 18,295 stars and an actively pushed repository on GitHub

Cons

  • macOS only today - Windows is announced as coming soon
  • No managed cloud version or team admin console; each user configures connectors and keys
  • Security coworkers are the most developed roles, while marketing, sales and EA are newer
  • Costs are invisible by design, so finance needs its own model-spend policy

Pricing Plans

Popular

OpenWorker app

$0

The complete MIT-licensed desktop app. The site states there is no paid tier and no subscription.


  • macOS desktop app (Windows coming soon)
  • Role-based coworkers plus custom ones
  • Approval check-ins for consequential actions
  • Local secret store for tokens
  • Connectors to Slack, email, calendar, CRM, docs and dev tools
Download for Mac

Bring your own model

Your provider bill

OpenWorker charges nothing for model usage - you connect your own keys and pay the provider directly.


  • OpenAI, Anthropic, Google API keys
  • Open-weight providers: Kimi, GLM, DeepSeek, Qwen, MiniMax, Mistral
  • Together AI and Fireworks
  • Fully local models via Ollama
  • Per-task model switching
See supported providers

Optional account

$0

Not required - everything works signed out. A free account exists so vendors such as Slack can authorise the connection flow.


  • One-click connector authorisation
  • Tokens handed to your machine, not stored in the cloud
  • No account needed for local models or manual connections
Create free account

Frequently Asked Questions

Editorial Review

4.4
ET

Editorial Team

September 27, 2026

Verified from openworker.com (homepage, FAQ, coworkers section) and the GitHub API on 27 September 2026: the app is free under an MIT licence with no paid tier, model access is bring-your-own-key across OpenAI, Anthropic, Google, open-weight providers and Ollama, connectors listed are Slack, Gmail, Outlook, Google Calendar, Notion, HubSpot, Attio, Google Drive, Dropbox, Box, GitHub, Jira, Linear and Asana, and macOS is available with Windows marked coming soon. The GitHub API reported github.com/andrewyng/openworker at 18,295 stars, 2,585 forks, MIT licence, created 20 July 2026 and last pushed 24 September 2026. topai.tools lists OpenWorker under AI Agents with open-source availability and marks it among the fastest-growing agents in the directory.

ET

Editorial Team

September 27, 2026

Editorial verdict: OpenWorker is the cleanest open-source answer we have seen to the seat-priced AI coworker category, because it inverts the usual deal - you keep the data and the model choice, and the vendor is not in the billing path between you and your provider. For a marketing ops lead the concrete value is the Marketing and Sales coworkers plus Slack-mention and schedule triggers: a Monday spend and attribution report, account research and follow-ups, and inbox triage, all inside HubSpot, Gmail and Notion where the work already lives. Our reservations are the ones we would tell a friend: it is a Mac-only desktop app today, security is clearly ahead of marketing in maturity, and there is no admin console or spend reporting, so teams have to govern model usage themselves. Buy it for the workflow control, not for hand-holding.

Related AI Agents

Have an agent to share?

Submit your AI marketing agent to reach thousands of marketers, agencies, and founders every month.

Submit Your Agent