Agent Barn

Agent Barn is a self-hosted control plane that lets teams create, run, and monitor AI agents across Slack, Teams, and Telegram from their own Kubernetes infrastructure.

Agent Barn is a self-hosted control plane for creating, configuring, and monitoring AI agents that work inside the chat tools teams already use. Rather than routing conversations through a vendor's cloud, the platform deploys into a customer's own Kubernetes cluster via Helm, with PostgreSQL holding product state and LiteLLM handling model routing and cost attribution. Agents are reachable through Slack, Microsoft Teams, and Telegram, so staff can ask them to search Jira, check GitHub activity, pull Confluence pages, or produce a scheduled report without leaving a normal conversation. The product is built for organizations with repeatable, high context workflows rather than one off chat experiments. Agent Barn highlights manufacturing coordination (RFQ intake, custom quoting, drawing checks, order status) and legal operations (procurement claims, precedent search, document review, email intake) as applied pilot areas, alongside general team ops tasks like weekly delivery reporting. Administrators define an agent's role, instructions, tools, and restrictions in versioned markdown templates, then attach reusable skills and encrypted credentials for services such as Jira, Confluence, GitHub, Bitbucket, and Gmail. In practice, a team starts by scoping a single workflow worth automating, deploys the control plane into its own infrastructure, and assigns each agent a runtime (OpenClaw or Hermes) along with the skills it needs. Every conversation, tool call, log line, lifecycle change, and cost figure is recorded centrally, giving operators a way to audit exactly what an agent did and why, broken down per agent, per tool, and per day. Among AI agent platforms, Agent Barn sits apart from fully managed SaaS offerings by keeping data, credentials, and infrastructure under the customer's control. That trade off suits engineering led organizations comfortable running Kubernetes workloads and willing to operate the platform themselves, more than teams looking for an instant, no-infrastructure agent sign up.