Skip to main content
TraceForge

TraceForge

Observe. Understand. Control.

A framework-agnostic Python library that forges raw AI-agent traces into structured, classified, risk-scored, and governance-assessed output.

SourceParserAdapterEnricherPipelineSink(s)

Framework-agnostic

Add a new agent framework with a single YAML mapping, no Python required. 22 mappings ship out of the box.

Runs anywhere

Runs from a laptop to CI. CPU-only, no heavyweight ML stack.

Classification & risk

Multi-dimensional classification with tree-sitter shell AST analysis and a 0–100 risk score mapped to MITRE ATT&CK.

Governance built in

Data labeling, information-flow control, drift and budget tracking, and rule-driven recommendations, with an opt-in gate layer.

Eight storage sinks

JSONL, SQLite, Parquet, S3, OTLP, webhook, console, and callback sinks, all configurable from YAML with error isolation.

Observation-first

Observes, enriches, and recommends by default. Enforcement is a separate, opt-in gate layer.