Use codemap find "SymbolName" to locate any symbol before reading — it gives exact file + line range and is faster than grep across the whole repo.

Building Agents with lauren-ai

Quick start

# agent.py — from __future__ import annotations IS allowed here (no @tool)
from __future__ import annotations

from lauren_ai import agent, use_tools
from .tools import get_price

@agent(model="claude-opus-4-6", system="You are a helpful shop assistant.")
@use_tools(get_price)
class ShopAgent: ...

Run it:

from lauren_ai import AgentRunnerBase, LLMConfig

cfg = LLMConfig(provider="anthropic", model="claude-opus-4-6")
runner = AgentRunnerBase(transport=transport)
result = await runner.run(ShopAgent(), "How much is a Widget?")
print(result.content)

Decorator order (mandatory)

Decorators are applied bottom-up in Python. Always write them top-to-bottom in this order so each decorator reads the metadata set by those below it:

@agent()           ← outermost (top)   — applied last, reads all metadata
@remember()        ← optional
@use_guardrails()  ← optional
@use_tools(...)    ← innermost (bottom) — applied first, sets USE_TOOLS_META
class MyAgent: ...

Full example with all optional layers:

from lauren_ai import agent, remember, use_guardrails, use_tools
from lauren_ai import PromptInjectionFilter, PIIRedactor
from .tools import search_database, send_email
from .memory import user_memory_store

@agent(model="claude-opus-4-6", system="You are a personal assistant.", max_turns=10)
@remember(store=user_memory_store, extract=True, inject=True, top_k=5)
@use_guardrails(
    input=[PromptInjectionFilter(), PIIRedactor()],
)
@use_tools(search_database, send_email)
class PersonalAssistant: ...

Safety invariants

AgentConfig parameters

All parameters are passed as keyword arguments to @agent():

Reference files