Framework Integrations

Prysm integrates natively with popular agent frameworks. Each integration captures framework-specific telemetry and optionally enables governance monitoring.

LangGraph

pip install prysmai[langgraph]

PrysmGraphMonitor provides graph-aware telemetry with node tracking, state transitions, and governance support.

from prysmai.integrations.langgraph import PrysmGraphMonitor

monitor = PrysmGraphMonitor(
    api_key="sk-prysm-...",
    governance=True,  # Enable behavioral detection
)

# Use as a callback handler in your LangGraph graph
graph = create_your_graph()
result = graph.invoke(
    {"messages": [("user", "Research quantum computing")]},
    config={"callbacks": [monitor]},
)

# Access governance results
if monitor.governance_session:
    report = monitor.governance_session.get_report()

What gets captured:

Event	Description
Node execution	Which graph nodes ran, in what order, with timing
State transitions	State changes between nodes
LLM calls	All LLM invocations with model, tokens, latency
Tool calls	Tool invocations with arguments and results
Conditional routing	Which edges were taken at decision points

Governance with LangGraph

When governance=True, the monitor automatically:

Starts a governance session when the graph begins
Reports all captured events to the behavioral detection engine
Runs a final behavioral check when the graph completes
Generates a governance report accessible via monitor.governance_session

CrewAI

pip install prysmai[crewai]

PrysmCrewMonitor captures crew execution telemetry including agent actions, task completions, and delegation events.

from prysmai.integrations.crewai import PrysmCrewMonitor

monitor = PrysmCrewMonitor(
    api_key="sk-prysm-...",
    governance=True,  # Enable behavioral detection
)

crew = Crew(
    agents=[researcher, writer],
    tasks=[research_task, write_task],
)

result = crew.kickoff(callbacks=[monitor])

What gets captured:

Event	Description
Agent execution	Which agents ran, with timing
Task completion	Task results and duration
Tool usage	Tools called by each agent
Delegation	When agents delegate work to other agents
LLM calls	All underlying LLM invocations

LlamaIndex

pip install prysmai[llamaindex]

PrysmSpanHandler integrates with LlamaIndex's instrumentation system.

from prysmai.integrations.llamaindex import PrysmSpanHandler
import llama_index.core

llama_index.core.global_handler = PrysmSpanHandler(
    api_key="sk-prysm-...",
)

# All LlamaIndex operations are now traced
index = VectorStoreIndex.from_documents(documents)
response = index.as_query_engine().query("What is quantum computing?")

All frameworks together: You can use multiple integrations simultaneously. Each one captures its own telemetry and sends it to the same Prysm project.

← Governance Dashboard →