Anter Product Demo — Building and Running Managed AI Agents
A full walkthrough of Anter, recorded from the live product. Watch a multi-agent system get built in the visual builder, then run end to end — triage routing, sub-agent delegation, MCP tool calls, and the execution trace and usage metrics behind every step.
2 min 32 sec · Recorded from the live product · Published
Chapters
What this demo shows
The recording opens in the agent workspace, where a customer-success system is assembled from named sub-agents — an intake-triage entry point, plus specialists for compliance, billing, refund risk, log investigation, and technical support. The routing graph makes the relationships explicit: solid edges are transfers, dashed edges are delegations.
From there the run moves across product surfaces in a single workflow. Triage picks a specialist path, delegation hands leaf work to a sub-agent, and MCP tools and HTTP tools are called where the agent needs data it does not hold. Each hop stays inspectable — the execution trace records which agent acted, why it routed the way it did, and what the tool returned.
The walkthrough closes on observability: token usage, model overrides, and execution targets shown in the same place the routing decisions were made, so the cost and behaviour of an agent are visible without leaving the builder.
Build the same thing yourself
Everything in this recording is available in the product — the visual builder, the routing graph, MCP tool imports, and the execution traces.