SHOWCALL

Open source agent observability

Know where your agents are.

It's 10 PM. Do you know where your agents are?

Showcall Agent Catalog makes AI agents visible in Backstage, across runtimes, clusters, A2A services, and the model gateway that powers them.

Explore the docs View on GitHub

v0.2.0 technical preview · Apache-2.0 · Backstage-native

DEMO FLEET / SAMPLE observed / live
Agent Catalog fleet view showing agents across multiple runtimes and clusters
kagentarka2aheuristic

The missing layer

Your service catalog knows your services. Now it can know your agents.

Agents are becoming ordinary production workloads, but they arrive through a growing collection of runtimes and frameworks. The result is a familiar platform problem: ownership is unclear, lifecycle is guessed, and usage lives somewhere else.

Agent Catalog observes what is actually running and turns it into ordinary Backstage entities that teams can search, own, govern, and understand.

Built for the messy middle

One catalog for agents that were never designed to share a catalog.

01 / DISCOVER

Find what exists

Discover kagent and ARK resources, labeled A2A Services, and unlabeled workloads that show signs of LLM use.

CRD · LABEL · HEURISTIC
02 / VERIFY

See what is alive

Fetch live A2A cards through the Kubernetes service proxy and preserve the last known truth when a cluster disappears.

CARD · REACHABILITY · LIFECYCLE
03 / GOVERN

Make usage legible

Connect agents to owners, models, tools, clusters, and LiteLLM usage so governance starts with evidence.

OWNER · DEPENDENCY · TRACTION

The agent card

Not a registry of intentions. A catalog of what runs.

Every discovered agent becomes a normal Backstage Component with a runtime, owner, lifecycle, reachability, and usage context. The result fits the platform workflows you already have.

See the entity model
Agent Catalog card on a Backstage entity page showing runtime, cluster, discovery, reachability, and request activity
Agent metadata stays next to the service and team context already in Backstage.

Start where you are

Try it locally. Read the model. Bring the hard questions.

01

Read the docs

Understand the discovery model, configuration, architecture, and security boundary.

→ 02

Run the demo

Install a real local runtime, populate a small agent estate, and open the fleet view.

→ 03

Talk to Showcall

Discuss multi-cluster rollout, runtime integration, governance, or enterprise support.

→