These are the executable counterparts to agentguard quickstart.
Each file is intentionally small:
- one supported stack
- one obvious install command
- one obvious run command
- local traces by default
They live in this repo for copy-paste onboarding and coding-agent setup. They are not included in the installed PyPI wheel.
| File | Stack | Local-first behavior |
|---|---|---|
agentguard_raw_quickstart.py |
Raw Python | Fully offline. No API keys or network calls. |
agentguard_openai_quickstart.py |
OpenAI | Local traces via local_only=True; OpenAI call still uses OPENAI_API_KEY. |
agentguard_anthropic_quickstart.py |
Anthropic | Local traces via local_only=True; Anthropic call still uses ANTHROPIC_API_KEY. |
agentguard_langchain_quickstart.py |
LangChain | Explicit callback wiring with local traces. |
agentguard_langgraph_quickstart.py |
LangGraph | Fully local graph example with node guards. |
agentguard_crewai_quickstart.py |
CrewAI | Explicit callback wiring with local traces. |
agentguard_pydantic_ai_quickstart.py |
Pydantic AI | Local traces around a Pydantic AI agent run using TestModel; no API keys or network calls. |
pip install agentguard47
agentguard doctor
agentguard quickstart --framework raw --write
python agentguard_raw_quickstart.py
agentguard report .agentguard/traces.jsonlThen move to the file that matches your real stack.
For Pydantic AI:
pip install agentguard47 pydantic-ai
python examples/starters/agentguard_pydantic_ai_quickstart.py
agentguard report .agentguard/traces.jsonl