Governance for every agent framework

Drop-in callback handler that instruments chains, LLMs, tools, agents, and retrievers. Enforcement runs on every chain entry and every tool call.

Tracing processor that hooks into the Agents SDK to capture agent/tool/generation spans and block disallowed tools before they run.

One call to instrument_crew() wires in step and task callbacks, chaining your existing callbacks rather than replacing them.

instrument_autogen_agent() wraps generate_reply so every conversation round runs through enforcement and every tool_calls / function_call entry is gated by tool-allowlist policies.

Dispatcher event handler for queries, LLM calls, retrievals, and agent tool calls. Enforces policies before tool execution and on the top-level query.

No-code workflow integration via HTTP Request nodes — templates for enforcement, approval polling, and trace ingestion.

instrument_openai() wraps the openai SDK client so every chat.completions.create enforces policies on the user prompt and records an LLM span with model + token usage.

Wrap an MCP ClientSession / Client so every callTool runs through tool-allowlist enforcement before the request reaches the MCP server.

Stream traces and policy violations to Splunk via HTTP Event Collector. Ships with an SPL saved-search pack so your SOC triages agent incidents in existing workflows.

Ship traces into a Log Analytics custom table via the Azure Monitor Logs Ingestion API, with ready-to-use KQL analytics rules that turn violations into Sentinel incidents.

Stream traces and governance context into Datadog Log Management via the Logs intake API, with monitor queries for violation bursts, cost runaway, and autonomy-tier signals.