Work groups · Open source · Self-hosted

Your team and its agents, in the same space

Create a shared workspace where people and agents collaborate with the same knowledge, connections, workflows and access rules. Everything stays under your organisation's control.

Create free account Explore the documentation

Or install it on your server:

curl -fsSL https://raw.githubusercontent.com/iagentshub/iAgents/main/install.sh | bash

Works with

What sets iAgents Hub apart

A work group is far more than a shared folder

It is a team's operating context. Resources keep their owner, members receive only the access they need, and everything can be used without copying settings or credentials.

Clear ownership and permissions for every resource.
Agents and workflows cascade-share the dependencies they need.
Connections remain private even when a group can use them.

See how groups work

Rubén can use María's connection without seeing or exporting her API key.

Your team's AI infrastructure

Move from isolated agents to a coordinated work group

Bring agents, people, models, knowledge, tools and workflows into a workspace with traceable ownership, permissions and usage.

Work groups

Each group is an operational workspace: it brings together members, agents, connections, LLM orchestrations, workflows, knowledge, prompts, skills and tools without duplicating them.

Multi-agent

Create as many specialised agents as you need, each with its own knowledge, its connections and its memory. No artificial limits.

Provider connections

Connect Claude, OpenAI, Gemini, Grok and Ollama with your own API keys. Switch providers without touching a line of your agents.

Knowledge

Turn documents, URLs and notes into real context for your agents, with no model retraining and no cloud dependency.

Self-hosted

Your data never leaves your server. Deploy on your own infrastructure with a single command, with no external dependencies.

Export to local tools

Export your agents to the native format of Claude Code, GitHub Copilot or OpenAI Assistants. Build once, use them wherever you work.

Explore the documentation

How it works

From zero to your first agent in three steps

01
Install
One command on your server. Pick Docker or a direct build from GitHub; the installer handles the rest.
02
Connect your keys
Add the API keys of the providers you already use. They stay in your instance and never travel anywhere.
03
Build your agents
Give them knowledge, memory and skills, share them with your team and export them to the tools you already work in.

Install and update

The same command installs and updates. On first install you can choose the deployment mode.

Linux terminal

curl -fsSL https://raw.githubusercontent.com/iagentshub/iAgents/main/install.sh | bash

Docker: pulls from GitHub Container Registry and enables automatic updates.

Without Docker: builds from GitHub and retains every feature.

In either mode you can install the full application, the backend only, or the frontend only.

Set up your first work group this afternoon

A free account on the managed hub, or one command if you would rather run your own server.

Create free account View the repository