agenwatch 0.1.0

Deterministic agent execution with runtime-enforced guardrails

AgenWatch

Deterministic agent execution with runtime-enforced guardrails.

AgenWatch is a production-grade Python SDK for running LLM agents with hard execution limits, budget enforcement, and deterministic replay — enforced during execution, not observed after.

---

Install

pip install agenwatch

With provider support:

pip install agenwatch[openai]     # OpenAI
pip install agenwatch[anthropic]  # Anthropic
pip install agenwatch[groq]       # Groq
pip install agenwatch[litellm]    # 100+ providers (Gemini, Mistral, etc.)
pip install agenwatch[all]        # All providers

---

Quick Example

from agenwatch import Agent, tool

@tool
def add(a: int, b: int) -> int:
    """Add two numbers."""
    return a + b

agent = Agent(
    name="calculator",
    model="gpt-4o-mini",
    tools=[add],
    budget=1.0,
)

result = agent.run("Add 2 and 3")

print(result.output)
print(result.cost)

---

Why AgenWatch

• Runtime-enforced budgets — kernel-level kill switches, not just logging
• No double-charging on retries — fingerprint-based deduplication
• Deterministic execution — replayable agent runs
• Provider agnostic — OpenAI, Anthropic, Groq, Gemini, and 100+ via LiteLLM

---

Status

v0.1.0 — stable core

---

License

MIT