PlumoAI

PlumoAI is an open-source platform that lets teams hire autonomous AI employees — not chatbots or simple automations, but role-based AI workers that plan their own approach, select agents from a library, build workflows, and report results back into a shared project board. The platform is currently in beta and available for self-hosting via Docker, with a managed cloud option also available.

What It Is

PlumoAI positions itself as an "AI Employee OS" — an operating system for autonomous AI workers that operate as real organizational units. Each AI employee is defined by six components: a role (e.g., Sales Manager, Research Analyst), tools (agents it can access), authorization (granular permissions on what it can read, write, or trigger), memory (short-term task context plus long-term persistent knowledge), presence (channels it operates on — email, Slack, calls, project board), and accountability (ownership of outcomes, not just task completion). The platform started as a project management and internal tooling product — a replacement for Jira and Confluence — and evolved into the AI employee layer on top of that foundation.

How AI Employees Work

When you assign a goal to an AI employee, it does not wait for you to configure a workflow. It reads the goal, scans the platform's agent library, selects the agents it needs, estimates token cost, builds the execution plan, and presents it for your approval before starting. For example, a goal like "book 50 enterprise meetings this month" causes the AI Sales Manager to autonomously pick a Data Scraping Agent, Workflow Builder Agent, Email Agent, Call Agent, CRM Agent, and Scheduler Agent — then run the full outreach motion and report results at month end. Human approval is required before execution begins, keeping humans in the loop at the plan stage.

Agent Ecosystem and Community

The platform ships with a growing library of agents that AI employees can reach for:

• Data Scraping Agent — LinkedIn, Apollo, web sources
• Email Agent — Gmail, Outlook, SMTP
• Call Agent — Calendly, Aircall, Twilio
• CRM Agent — HubSpot, Salesforce, Pipedrive, or PlumoAI's own CRM
• Workflow Builder Agent — sequences, triggers, logic
• Memory Agent — long-term context and recall
• OpenClaw — multi-step workflow execution engine for advanced reasoning
• Hermes Agent — self-improving with persistent skills

Anyone can build and submit an agent via GitHub (agents live in ai-agents/ in the repo). The platform discovers submitted agents automatically and makes them available to all AI employees. Pre-trained AI employee templates (Sales Manager, Research Manager, Operations Manager, Marketing Manager, Product Manager) are available out of the box; community-built employees can be downloaded as zip files and imported in one step.

Deployment Model and Self-Hosting

PlumoAI runs as a multi-container Docker Compose stack (UI + APIs + AI service + MySQL + MongoDB + Milvus) behind Traefik. The GitHub README documents installation for Windows, Linux, and macOS in both domain mode (HTTPS with Let's Encrypt) and localhost mode. System requirements for self-hosting are substantial: minimum 2 vCPU and 16 GB RAM, with 30 GB disk; production deployments are recommended at 4+ vCPU, 32+ GB RAM, and 100+ GB SSD. Self-hosting is free with no feature limits, no seat caps, and no per-action billing from PlumoAI — users pay only for their own infrastructure. The managed cloud option uses a prepaid credit wallet with usage-based billing per AI employee action.

Platform Scope and Availability

Beyond AI employees, PlumoAI includes a full project management workspace where humans and AI employees share the same board — tasks assigned to either show up identically, and AI employees log results, token usage, and status directly back into the project. A low-code/no-code builder is included for creating internal apps without engineering effort. The platform is accessible via web app, an Android app on Google Play (iOS listed as coming soon), and a desktop app download. The web app and mobile app can both connect to a self-hosted instance, keeping all data on the user's own infrastructure.

Update: v1.0.3

The latest release is v1.0.3, published in May 2026, with the repository last updated in August 2026. The project is currently in beta — the GitHub README explicitly notes beta status and directs users with setup issues to a Discord live support call. The roadmap is organized into three phases: Phase 1 (core platform and Docker deployment, complete), Phase 2 (AI agent developer framework and MCP-based integrations), and Phase 3 (pre-built AI employee templates for sales, research, operations, marketing, and product management). The repository has 38 stars and 20 forks as of the latest data.