AgentConnect

0 reviews

AgentConnect is an open-source platform for building, configuring, and operating teams of AI agents that work alongside people in the tools a company already uses. Each agent is defined in a web console with its own instructions, runtime, model, workspace, tools, memory, and machine. Agents run multi-step work autonomously, started from a chat message, an issue or pull request, a webhook, or a schedule. An agent can delegate bounded work to other agents, and each worker runs in its own session under its own permissions, returning a result rather than access. Agents reach external systems through MCP servers, through GitHub, GitLab, Gitea, and Linear, and through Slack, Telegram, Discord, and Lark. Reviewed organization Knowledge gives every agent a governed source of shared reference material. Operators get a Playground for testing agent behavior before deployment, a session view with transcripts and visibility controls, and usage and billing reporting. Recurring operations run on a schedule or webhook and bring exceptions into a shared conversation so the human decision stays visible. Five separate permission layers govern platform access, team visibility, which agents may call one another, who may read a session, and runtime sandboxing per agent. AgentConnect is licensed under Apache-2.0 and can be self-hosted with Docker Compose or Kubernetes for data residency, or used as a managed cloud service.