Service map

Auto-discovered topology of all services and agents communicating in your system — built from trace data with no manual configuration required.

The service map gives you a visual overview of how your services and AI agents connect to each other. It is built automatically from the span data you send to AgentVista — no configuration files, no manual topology definitions.

The service map requires trace data to exist. If you have just created your account or have not yet ingested any traces, the map will be empty. Start the quickstart to send your first trace.

How it works

Every time a span is ingested, AgentVista records its parent_span_id. When a span from service A has a child span in service B, that means A called B. The service map aggregates all of these parent-child relationships across all your traces and renders them as a directed graph:

• Nodes represent individual services and agents
• Edges represent call relationships, directed from caller to callee
• Edge weight is the number of spans that crossed that relationship (call volume)

The map refreshes automatically as new trace data flows in.

Node types

Every node has a node_type derived from the span type that produced it:

Node type	Produced by span type	Examples
agent	agent spans	Your AI agents instrumented with the AgentVista SDK
service	http spans	REST APIs, microservices, internal HTTP services
database	db spans	PostgreSQL, Redis, any database query span
external	custom spans	Third-party APIs, external services

AI agents appear as first-class nodes on the map, placed alongside your infrastructure services. A single map can show you an agent calling an internal API, which in turn queries a database — the full call chain in one view.

Time range filtering

By default, the map shows topology derived from traces in the last 24 hours. You can change this using the hours query parameter:

# Last 24 hours (default)
GET /api/v1/dashboard/service-map/?hours=24

# Last 7 days
GET /api/v1/dashboard/service-map/?hours=168

# All time
GET /api/v1/dashboard/service-map/?hours=0

The dashboard UI provides a time range picker that maps to the same parameter.

API response format

{
  "nodes": [
    {
      "id": "support-agent",
      "name": "support-agent",
      "node_type": "agent",
      "span_type": "agent",
      "span_count": 1240
    },
    {
      "id": "user-service",
      "name": "user-service",
      "node_type": "service",
      "span_type": "http",
      "span_count": 3820
    },
    {
      "id": "postgres-main",
      "name": "postgres-main",
      "node_type": "database",
      "span_type": "db",
      "span_count": 9100
    }
  ],
  "edges": [
    {
      "source": "support-agent",
      "target": "user-service",
      "weight": 1240
    },
    {
      "source": "user-service",
      "target": "postgres-main",
      "weight": 3820
    }
  ]
}

Each node’s id and name match the service name or agent name used in the span data. The span_count tells you how active that node was in the selected time window. Edge weight reflects the number of spans observed crossing that relationship.

Common use cases

Dependency discovery

Ask “which services does my agent depend on?” — the map shows you every downstream dependency your agent touches, including any you may have missed during development.

Traffic flow analysis

See how traffic flows through your system. Identify which services carry the most call volume and which are potential bottlenecks.

Unexpected connections

Spot call relationships you did not expect — for example, an agent calling a service it should not, or a service making calls to an external API that bypasses your intended architecture.

Incident scoping

When an agent starts failing, check the service map to quickly identify which downstream services are in its call path and narrow down where the problem may originate.