Multi-Agent

Spawn isolated sub-agents with different models, system prompts, and tool sets.

---

How It Works

flowchart TD
    A["Parent Agent (DevDuck)<br/>Model: Bedrock Claude"] --> B["use_agent()<br/>model: anthropic<br/>tools: file_write"]
    A --> C["use_agent()<br/>model: openai<br/>tools: calculator"]
    A --> D["use_agent()<br/>model: ollama<br/>tools: shell"]
    B --> E["Isolated context<br/>Own messages + tools"]
    C --> F["Isolated context<br/>Own messages + tools"]
    D --> G["Isolated context<br/>Own messages + tools"]

Each sub-agent runs in its own isolated context with:

• Different model provider — mix cloud + local
• Custom system prompt — specialized personality
• Limited tools — only what's needed (security)
• Own message history — no context bleed

---

use_agent Tool

Parameter	Type	Description
prompt	string (required)	The task for the sub-agent
system_prompt	string (required)	Custom personality/instructions
model_provider	string	bedrock, anthropic, openai, ollama, env, etc.
model_settings	object	{"model_id": "...", "params": {...}}
tools	array	Tool names to make available

---

Examples

Different Model for Creative Task

use_agent(
    prompt="Write a haiku about artificial intelligence",
    system_prompt="You are a minimalist poet who writes profound haikus.",
    model_provider="anthropic"
)

Local Model for Privacy

use_agent(
    prompt="Summarize this confidential document",
    system_prompt="You summarize documents concisely. Never output raw data.",
    model_provider="ollama",
    model_settings={"model_id": "qwen3:8b"}
)

Tool Isolation

# Sub-agent can only read files, not write or execute
use_agent(
    prompt="Review this code for security issues",
    system_prompt="You are a security auditor.",
    model_provider="bedrock",
    tools=["file_read"]
)

Model Comparison

# Get answers from multiple models
for provider in ["anthropic", "openai", "bedrock"]:
    use_agent(
        prompt="Explain quantum computing in one sentence",
        system_prompt="You are concise and precise.",
        model_provider=provider
    )

Specialized Workflows

# Step 1: Research with web access
use_agent(
    prompt="Find the latest Python 3.13 features",
    system_prompt="You are a research assistant.",
    tools=["shell", "scraper"]
)

# Step 2: Code with file access
use_agent(
    prompt="Write a demo using the new features",
    system_prompt="You are a Python expert.",
    tools=["file_write", "editor", "shell"]
)

---

Zenoh P2P Multi-Agent

For distributed multi-agent coordination across machines, see Zenoh P2P.

# Broadcast a task to all connected DevDuck instances
zenoh_peer(action="broadcast", message="run linting on your current project")

# Send to a specific peer
zenoh_peer(action="send", peer_id="hostname-abc123", message="deploy to staging")