Skip to content
Agent HangarAgent Hangar

What's Agent Hangar?

One home for every AI agent your company builds, whatever tool built it. Build agents where you already work, run each one in its own container, and use them from any platform.
InstallConceptsInteractive demo

People create agents where they already work: Claude, ChatGPT, Cursor, VS Code, Codex, OpenCode. Agent Hangar brings them all to one place, runs each one in its own container, and makes them usable from any platform: OpenAI-compatible, MCP, A2A and ACP, and inside your own app over AG-UI. Everyone can find them, talk to them, and combine them into new agents.

Why Agent Hangar

Section titled “Why Agent Hangar”

Built where you already work

Ask Claude, ChatGPT, Codex or Cursor for an agent. It is built, tested and shipped through the Hangar MCP, with no new tool to learn.

Usable from anywhere

Every agent speaks OpenAI-compatible, MCP, A2A and ACP, and runs inside your own app over AG-UI.

Governed by default

Teams, access requests, approvals, authority rules for Digital employees, tests before production, DLP and an audit trail.

Cost under control

Usage and cost per agent, budgets, the right reasoning effort per role, long-context compaction and a cheaper model for simple requests.

Quick start

Section titled “Quick start”
Terminal window
git clone https://github.com/valteresj2/agent-hangar && cd agent-hangar
./scripts/install.sh

The guided installer writes .env, starts the stack, registers and tests your LLM, creates the admin and generates the configuration of your AI tools. hangar doctor checks the result. See Installing Agent Hangar for the requirements, a cloud VM or Kubernetes.

Building blocks

Section titled “Building blocks”
Agent specWhat an agent is: instructions, model, skills, tools, MCPs, sub-agents and tests.
Reuse before you buildNew agents from the pieces of existing ones, without changing them.
Execution profileReasoning effort, output limit, long-context compaction and the model per response.
HarnessesClaude Code, Codex, Hermes and DeepSeek Harness, one container per task.
Connectors (MCP)Official servers, the Docker MCP catalog and the GitHub MCP Registry, through one bridge.
Digital employeesAgents with a role, a manager, authority rules, a career and an org chart.
Connecting clientsClaude, ChatGPT, Cursor, VS Code, LibreChat, your app over AG-UI.
Security modelWhat is isolated, what is checked, and the limits we document.

Run it in production

Section titled “Run it in production”
KubernetesHelm chart for GKE, AKS and EKS, with a kernel sandbox for agents.
ObservabilityPrometheus metrics, OpenTelemetry traces and logs.
Backup and upgradehangar backup, restore and upgrade, and the chart hooks.
HTTP APIEvery endpoint of the central.