These examples mirror the three main ways to use PrysmAI:

1. Proxy path for application traffic
2. MCP path for agent runtimes
3. Unified session path for correlated runs across requests and governance evidence

They also include framework-oriented examples for developers who want Prysm to plug into an orchestration layer instead of building telemetry wiring by hand.

Set your Prysm credentials:

export PRYSM_API_KEY="sk-prysm-..."
export PRYSM_BASE_URL="http://localhost:3000/api/v1"

Then run any example:

python examples/proxy_chat.py
python examples/mcp_agent_runtime.py
python examples/unified_session.py
python examples/langgraph_monitor.py
python examples/agent_framework_monitor.py

• proxy_chat.py

  • Smallest application example
  • Routes an OpenAI-style request through Prysm

• mcp_agent_runtime.py

  • MCP-oriented example
  • Shows how to derive MCP connection config and record external runtime events

• unified_session.py

  • Best example of the current Prysm product shape
  • Correlates model traffic, decisions, tools, and file changes under one run

• langgraph_monitor.py

  • LangGraph callback example
  • Starts governance, wires Prysm as a graph callback, and closes cleanly

• agent_framework_monitor.py

  • Agent Framework middleware example
  • Shows where Prysm wraps agent execution, tool activity, and chat-model calls