Agents and Sessions

Agents

An agent in Maev is a named entity that maps to a specific AI agent in your codebase. Agents are created automatically the first time telemetry is received with a given name.

You set the agent name with the agent_name parameter:

maev.init(api_key="vl_xxx", agent_name="Customer Support Agent")

If you omit agent_name, it defaults to "default". Give your agents descriptive names so they are easy to identify in the dashboard.

Multiple agents

If you run several different agents in your application, initialize each one separately with a distinct name:

# In your support agent script
maev.init(api_key="vl_xxx", agent_name="Support Agent")
 
# In your sales agent script
maev.init(api_key="vl_xxx", agent_name="Sales Agent")

Each agent gets its own page in the dashboard with its own session history, failure stats, and cost breakdown.

Sessions

A session represents one execution of your agent from start to finish. Every time you run your agent script, a new session is created.

Session lifecycle

What a session contains

• Full timeline of every LLM call, tool call, and error
• Total cost for the session
• Total duration
• Failure classification (if any)
• Inspector results (if inspectors are enabled)

Session timeline

Each event in a session has a step number, type, and payload. The timeline shows you exactly what your agent did in order, making it easy to trace back to the root cause of a failure.