Skip to content

Introduction

Workforce0 is an open-source software package that behaves like an AI product team. You install it once. After that it can:

  • Listen in on meetings (Google Meet / Zoom recordings, uploads, or a phone dial-in) and transcribe them with domain-aware prompts.
  • Write briefs (the equivalent of a PM’s one-pager) from those transcripts, with follow-up questions it posts back to the exec.
  • Decompose an approved brief into a plan — 1 to 6 steps — with each step assigned to a specialist AI role (BA, architect, dev, QA).
  • Dispatch those steps as tickets into a queue. Specialist agents pull them, do the work, and report back.
  • Summarise status in Slack / WhatsApp / Teams so the exec stays in the loop without ever opening the UI.
  • Keep the loop tight: failed tickets get replanned automatically, with the failure signature fed back into the plan prompt.

Everything runs on your infrastructure. Every LLM call uses your API keys. Every database row lives in your Postgres.

Workforce0 is deliberately not:

  • A SaaS. We don’t host anything. No multi-tenant cluster to share with strangers. No monthly fee.
  • A chat UI. The exec doesn’t prompt Claude in a web app. They approve a brief, answer a clarifying question, or redirect a step — always by replying to a message.
  • A fully-autonomous “agent swarm.” Humans stay on the critical path. Plans are drafted, not executed silently. Work that affects shared systems waits for a human approval.
  • A no-code builder. We optimize for developer teams that can run Docker and understand API keys. The exec gets a curated product; the installer gets YAML and logs.

Workforce0 fits teams where:

  1. There’s an exec or operator with a lot of meeting-driven, decision-heavy work — VP Product, CPO, Head of Ops, founding CEO.
  2. There’s at least one technical person (founding engineer, tech lead, platform engineer) willing to run the installer persona.
  3. The team uses Slack / Teams / WhatsApp / Google Chat as the daily communication surface.
  4. The team has at least one API key for a frontier model — Anthropic, OpenAI, or Google (a free-tier Gemini key is fine to start).

If that’s not you, that’s fine — there are a lot of good AI tools. But Workforce0 won’t give you value in isolation.

We take the split-audience separation seriously. You’ll see it everywhere:

SurfaceSpeaks to
This docs siteInstaller
The repo’s README / CLI errorsInstaller
The setup wizard in the web UIInstaller
Slack / WhatsApp / Teams messagesExec (consumer)
The web UI dashboard / audit logBoth

If you find a page, a message, or a log line that blurs these two audiences, open an issue. It’s a bug.

  • If you have 5 minutes and want to see it work: Quickstart.
  • If you need a production-grade install plan: Installation.
  • If you want to understand the pieces before installing anything: System overview.