Worklone is a cloud-managed AI employee platform for organizations. It lets you create, manage, and deploy AI employees for your business. These are not static chatbots: they are autonomous agents that reason, use tools, execute workflows, collaborate with each other, and get smarter over time through self-learning.
Think of Worklone as your digital workforce: hire AI employees for product management, engineering, sales, support, operations, or any role you define. Each employee learns your preferences, builds skills from experience, and improves with every interaction.
This repository is released under the GNU Affero General Public License v3.0 (AGPL-3.0). If you use, modify, or host this software over a network, you must make the entire source code of your modified version publicly available under the same license. For commercial use without open-sourcing your code, please contact us for a commercial license. See LICENSE.
| Traditional Automation | Worklone AI Employees |
|---|---|
| Rigid if-then rules | Autonomous reasoning with ReAct |
| Manual updates required | Self-learning from experience |
| Single-purpose bots | Multi-tool, multi-skill agents |
| No memory or context | Remembers users, builds skills |
| Closed-source, expensive | Research-use source available, self-hosted |
- Learns from every interaction — automatically captures user preferences, work styles, and domain knowledge
- Builds skills autonomously — discovers multi-step procedures through trial-and-error and saves them as reusable skills
- Improves over time — the more you use it, the better it gets at understanding your needs
- Create custom employees with specific roles, personalities, and capabilities
- Self-learning employees — adaptive employees that improve through production feedback loops
- Assign 500+ tools — from GitHub and Slack to Salesforce and Stripe
- Track performance — monitor tokens, costs, and activity per employee
- Team collaboration — employees can message each other and work together on tasks
- 500+ pre-built tools across GitHub, Slack, Gmail, Jira, Notion, Salesforce, Stripe, Linear, HubSpot, Google Drive, Google Calendar, and more
- Custom tool framework — add your own tools in under 50 lines of code
- Native function calling — agents autonomously decide which tools to use and when
- Visual workflow builder — 13 block types including agents, tools, conditions, loops, and parallel execution
- Multiple triggers — API, webhook, schedule (cron), or manual
- Human-in-the-loop — approval blocks for critical decisions
- Background execution — long-running workflows with monitoring and history
- 100% self-hosted — your data never leaves your infrastructure
- Persistent data layer — SQLite for application state, Redis for dispatch/realtime coordination
- Multi-tenant — owner-based data isolation with session and API key auth
- Docker Desktop (or Docker Engine + Compose)
- An OpenRouter API key (or NVIDIA API key)
git clone https://github.com/hritvikgupta/worklone.git
cd worklone
cp .env.example .envEdit .env and set:
OPENROUTER_API_KEY=...DEPLOYMENT_MODE=self_hosted(for open-source/self-hosted usage)
./scripts/docker-up.shThe script:
- starts Docker if needed
- frees frontend port conflicts
- builds images
- starts
frontend + backend + redis
Stop services:
docker compose downOpen:
- Frontend:
http://localhost:3000 - Backend API:
http://localhost:8000 - Backend docs:
http://localhost:8000/docs
| Document | Description |
|---|---|
| Architecture | System design, data flow, and principles |
| AI Employees | How employees work, create, and collaborate |
| Self-Learning | How employees learn and improve over time |
| Tool System | Available tools and how to build custom ones |
| Workflows | DAG workflow engine, block types, and triggers |
| API Reference | REST endpoints, WebSocket, and authentication |
| Contributing | How to contribute, style guide, and testing |
| Security | Security policy and reporting |
Self-learning product employees can:
- Write PRDs and user stories
- Prioritize features using RICE or MoSCoW
- Plan sprints and roadmaps
- Analyze user feedback and metrics
AI engineers can:
- Create and manage GitHub issues and PRs
- Review code and suggest improvements
- Run tests and debug failures
- Document codebases and APIs
AI sales reps can:
- Manage Salesforce leads and opportunities
- Draft personalized outreach emails
- Track deal pipelines and forecast revenue
- Schedule meetings via Google Calendar
AI support agents can:
- Respond to tickets via Slack or email
- Search knowledge bases (Notion, Google Drive)
- Escalate complex issues to human agents
- Track satisfaction and resolution metrics
AI ops managers can:
- Monitor Stripe subscriptions and payments
- Generate reports from multiple data sources
- Automate recurring tasks with scheduled workflows
- Coordinate across teams and tools
Backend: FastAPI, Python 3.11 (Docker image), SQLite, Redis, httpx, Pydantic, Uvicorn Frontend: React 19, TypeScript, Vite, Tailwind CSS v4, shadcn/ui, Recharts LLM Providers: OpenRouter, OpenAI, Groq, NVIDIA NIM (user-configurable) Integrations: GitHub, Slack, Gmail, Jira, Notion, Salesforce, Stripe, Linear, HubSpot, Google Drive, Google Calendar, and more
- Autonomous Skill Acquisition (learning new tools via documentation)
- Multi-Agent Collaboration Framework
- One-Click Docker Deployment
- Enterprise SSO & RBAC
- Vector memory for long-term knowledge retention
- Employee performance analytics dashboard
- Marketplace for pre-built employee templates
We welcome contributions! Whether it's fixing a bug, adding a tool, or improving documentation—every contribution matters. See our Contributing Guide to get started.
This project is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0).
This is a strong copyleft license that ensures the software remains free and open-source. If you modify this software and host it over a network (like a SaaS), the AGPL requires you to make your modified source code available to your users under the same license.
Commercial use of any kind — including reselling, hosting as a paid service, or incorporating into a commercial product without open-sourcing your own code — requires a separate commercial license. Contact the maintainers for commercial licensing.
Built with ❤️ by the Worklone Community