Multi-Agent Output

Control how a team of agents displays its work — silent, minimal, or verbose

Multi-agent output controls how an AgentTeam displays its progress and results. Streaming is on by default for responsive output — disable it explicitly if you need compatibility with sync-only providers.

from praisonaiagents import Agent, AgentTeam

researcher = Agent(name="Researcher", instructions="Research topics")
writer = Agent(name="Writer", instructions="Write content")

team = AgentTeam(agents=[researcher, writer])
team.start("Research and write about AI")

The user runs a team; output mode controls whether progress stays silent, minimal, or verbose.

Quick Start

1. Default Silent Mode

   from praisonaiagents import Agent, AgentTeam
   
   researcher = Agent(name="Researcher", instructions="Research topics")
   writer = Agent(name="Writer", instructions="Write content")
   
   team = AgentTeam(agents=[researcher, writer])
   result = team.start("Research and write about AI")  # Silent by default
   
2. Verbose Output

   team = AgentTeam(
       agents=[researcher, writer], 
       output="verbose"
   )
   team.start("Research and write about AI")  # Rich display
   
3. Streaming Output

   from praisonaiagents import AgentTeam, MultiAgentOutputConfig
   
   team = AgentTeam(
       agents=[researcher, writer],
       output=MultiAgentOutputConfig(verbose=2, stream=True)
   )
   team.start("Research and write about AI")  # Verbose + streaming
   

---Multi-agent output controls how an AgentTeam displays its progress and results.

from praisonaiagents import Agent, AgentTeam

researcher = Agent(name="Researcher", instructions="Research topics")
writer = Agent(name="Writer", instructions="Write content")
team = AgentTeam(agents=[researcher, writer], output="verbose")
team.start("Research and write about AI")

The user picks silent, minimal, or verbose output; the team streams progress accordingly.

---

Presets

Preset	verbose	stream	Use case
"silent"	0	False	Programmatic / API / batch
"minimal"	1	True	Short status lines
"verbose"	2	True	Rich panels, full trace (default)

Streaming is ON by default in "minimal" and "verbose" presets. Only "silent" disables streaming. If you need compatibility with sync-only providers, explicitly set stream=False with output=["verbose", {"stream": False}] or output=MultiAgentOutputConfig(verbose=2, stream=False).

---

Configuration Options

MultiAgentOutputConfig API Reference

Full API reference

Option	Type	Default	Description
verbose	int	0	Verbosity level (0=silent, 1=minimal, 2+=verbose)
stream	bool	True	Stream tokens as the team’s agents respond

---

Disabling streaming for a multi-agent team

from praisonaiagents import Agent, AgentTeam, MultiAgentOutputConfig

team = AgentTeam(
    agents=[researcher, writer],
    output=MultiAgentOutputConfig(verbose=2, stream=False),
)
team.start("Research and write about quantum computing")

Or via the array form (preset + override):

team = AgentTeam(
    agents=[researcher, writer],
    output=["verbose", {"stream": False}],
)

Streaming is enabled by default. If you’re using sync-only providers like the sync OpenAI adapter or Deepseek (sync), explicitly set stream=False to ensure compatibility.

---

Choosing the right preset

---

Common Patterns

Pattern 1: Programmatic team (default)

from praisonaiagents import Agent, AgentTeam

team = AgentTeam(agents=[researcher, writer])
result = team.start("Topic")  # Returns result, no display

Pattern 2: Verbose terminal demo

team = AgentTeam(agents=[researcher, writer], output="verbose")
team.start("Topic")

Pattern 3: Verbose + streaming (provider must support sync stream)

team = AgentTeam(
    agents=[researcher, writer],
    output=["verbose", {"stream": True}],
)
team.start("Topic")

---

Best Practices

Use 'silent' for production APIs

output="silent" (the default) has zero display overhead and works on every provider.

Test streaming with single agent first

Sync streaming is not supported by every provider. Verify with a single agent before enabling for teams.

Choose 'verbose' for demos, 'minimal' for logs

verbose is great for live terminal walkthroughs; minimal keeps log files readable.

Use async for broader streaming support

If you use astart() (async), streaming is more widely supported because async adapters handle sync-only providers via background threads.

---

Related

Single-agent streaming (auto-detect + fallback)

Single-agent output configuration

Sequential, parallel, hierarchical patterns

Auto-generated SDK reference