Skip to main content

ClaudeAgent

What it is​

A thin wrapper around IntentAgent that registers a “Claude” agent with a predefined system prompt and a set of intents routed to a model callable.

Public API​

  • create_agent(agent_shared_state: Optional[AgentSharedState] = None, agent_configuration: Optional[AgentConfiguration] = None) -> IntentAgent

    • Builds and returns a configured ClaudeAgent.
    • Wires:
      • chat_model from naas_abi_marketplace.ai.claude.models.claude_sonnet_3_7
      • a system prompt (SYSTEM_PROMPT) with tool listing injected (tools list is empty in this module)
      • a default AgentSharedState(thread_id="0") if not provided
      • a default AgentConfiguration(system_prompt=...) if not provided
      • a set of intents that target call_model for analysis/writing/code tasks

  • class ClaudeAgent(IntentAgent)

    • No additional behavior; inherits everything from IntentAgent.

Configuration/Dependencies​

  • Depends on naas_abi_core.services.agent.IntentAgent for:
    • IntentAgent, AgentConfiguration, AgentSharedState, Intent, IntentType
  • Imports the chat model as:
    • from naas_abi_marketplace.ai.claude.models.claude_sonnet_3_7 import model
  • System prompt text indicates usage of an Anthropic API key via environment, but this module does not read env vars directly.

Usage​

from naas_abi_marketplace.ai.claude.agents.ClaudeAgent import create_agent
from naas_abi_core.services.agent.IntentAgent import AgentSharedState

agent = create_agent(agent_shared_state=AgentSharedState(thread_id="123"))
# Use `agent` via the IntentAgent interface provided by naas_abi_core.

Caveats​

  • This module defines no tools (tools = []), so the system prompt’s tool list will be empty.
  • ClaudeAgent itself adds no methods; behavior is entirely determined by the upstream IntentAgent implementation and the imported model.