Agent

agent

Agent model — a participant (node) in a Claw society graph.

Agents can be AI-powered (backed by an LLM) or human (interfacing via CLI, Slack, GitHub, or webhook). An agent is unaware of its position in the graph topology — it only receives events along its edges.

Example::

from claw import Agent

coder = Agent(
    "coder",
    role="implementer",
    model="claude-opus",
    instructions="Write clean, tested Python code.",
    tools=[Tool(name="file_edit", description="Edit a file")],
)

human = Agent("maintainer", role="maintainer", human=True, interface="cli")

Tool

Bases: BaseModel

A tool/action available to an agent.

Parameters:

Name	Type	Description	Default
name		Unique tool identifier (e.g. "file_edit", "shell_exec").	required
description		Human-readable description of what the tool does.	required
parameters		JSON Schema defining the tool's input parameters.	required

Agent

Bases: BaseModel

A participant in a Claw society.

Agents are graph nodes. The same agent definition produces different behavior depending on its edge context — the compiler generates per-agent prompts from graph topology.

Parameters:

Name	Type	Description	Default
name		Unique identifier for the agent within a society.	required
role		Semantic role label (e.g. "implementer", "critic", "dispatcher").	required
model		LLM model identifier. Required for AI agents, None for humans.	required
instructions		Agent-specific instructions included in compiled prompts.	required
tools		List of tools available to this agent.	required
human		Whether this is a human agent (different runtime interface).	required
interface		Interface type for human agents ("cli", "github", "slack", "webhook").	required

Raises:

Type	Description
ValueError	If a human agent lacks an interface or an AI agent lacks a model.