Multi-tool research agent

Bind Brime's three tools to a LangChain agent that picks the right one per question.

Wire the three Brime tools (BrimeSearch, BrimeExtract, BrimeResearch) into create_agent and let the LLM pick which to call.

from langchain.agents import create_agent
from langchain_brime import BrimeSearch, BrimeExtract, BrimeResearch

agent = create_agent(
    model="anthropic:claude-3-5-haiku-latest",
    tools=[BrimeSearch(), BrimeExtract(), BrimeResearch()],
    system_prompt=(
        "You're a research assistant. Use brime_search for quick lookups, "
        "brime_extract for full page content, and brime_research only when "
        "the user asks for a synthesised report."
    ),
)

result = agent.invoke(
    {"messages": [{"role": "user", "content": "What new features shipped in the latest Stripe API release?"}]}
)
print(result["messages"][-1].content)

The agent will (typically) call brime_search once, optionally brime_extract for the deepest source, and synthesise a final answer with citations.

For how deep research runs end to end, see How research works.

This recipe will be expanded with full traces, LangSmith integration, and prompt tuning notes.