downy
OpenClaw Alternative - Create a team of agents and work with them from anywhere.
What is it?
What it is
Downy is a self-hosted platform for creating and interacting with a team of AI agents, each with its own personality, skills, tools, and workspace, accessible from any device via a purpose-built UI.
Why it exists
To provide the best user experience for working with multiple agents without relying on CLI or external tools, while allowing users to use any model (including OpenAI subscriptions) and run the service securely on their own Cloudflare account.
Who should use it
Developers, teams managing AI agents, Cloudflare users, and those needing self-hosted multi-agent systems with customizable agents.
Who should avoid it
Non-technical users, those avoiding self-hosted solutions, or users uncomfortable with security risks like prompt injection.
How it works
A quick walkthrough in plain English
How downy works
Step 1 of 3
You interact with it
Open downy, send a request, or connect it to your stack.
Features
Advantages
- Full control over data and infrastructure
- User-friendly interface for managing multiple agents
- Cost-effective model usage with flat-rate options
- Secure access through Cloudflare's network boundary
- Flexible model integration without API key management
- Centralized workspace management without external tools
- Scalable architecture for agent-based workflows
Disadvantages
- Requires Node 24 LTS and specific tooling (pnpm, Alchemy)
- Complex setup process involving Cloudflare and Exa API
- Risk of prompt injection due to self-editing features
- Dependency on Exa API key with potential usage limits
- Limited topic categorization in metadata
- Potential learning curve for Cloudflare Access configuration
- Model performance may vary based on selected provider
Installation
cloud
1. Clone repo and install dependencies: `git clone https://github.com/bensenescu/downy && cd downy && pnpm install` 2. Set up env vars: `cp .env.example .env` and fill in EXA_API_KEY, ALCHEMY_PASSWORD 3. Deploy via Alchemy: `pnpm deploy` after configuring Cloudflare Access in the dashboard
FAQ
What makes Downy different from other multi-agent platforms?
Downy features a purpose-built UX where each agent has its own unique personality, skills, tools, and dedicated workspace. Unlike other tools, you can manage workspaces, tools, and background tasks directly within the app without needing a CLI or Obsidian.
Which AI models can I use with Downy?
Downy is highly flexible. It uses Kimi 2.6 on Workers AI by default (requiring a Cloudflare Workers Paid plan), but you can also connect to any model via OpenRouter or use your existing ChatGPT Plus/Pro subscription via a Pi proxy.
How is my Downy instance secured?
Downy is designed to be self-hosted behind Cloudflare Access. This ensures that all traffic is gated by Cloudflare's authentication layer, and the Worker will reject any request that hasn't been authenticated via your configured Cloudflare team domain.
What are the hardware and software requirements for deployment?
You will need Node 24 LTS, pnpm, a Cloudflare account, and an Exa API key for search capabilities. While the free Workers plan works for some models, using Workers AI requires the Cloudflare Workers Paid plan ($5/mo).
Is it safe to give Downy access to my tools and data?
Because Downy features powerful self-editing capabilities, it is susceptible to prompt injection. You should be cautious and considerate about the specific data and tools you grant the agents access to, as the project is currently in rapid development.
Featured in Videos
YouTube tutorials and walkthroughs for downy
Alternatives
Similar projects ranked by category, topics, and text overlap.
