Skip to content

GroundGuard

GroundGuard provides deterministic assertions for AI agent answers. It checks whether the final answer uses the facts that tools already returned, without calling a second LLM judge.

The core loop is small:

  1. Register tool facts in a local Ledger.
  2. Extract numeric claims from the final answer.
  3. Match claims against the ledger.
  4. Apply a policy before the answer is released.

GroundGuard is not a tracing dashboard, hosted observability service, or LLM-as-judge evaluator. It is the narrow gate that catches "the tool had the number, but the model ignored or rewrote it."

Start Here

from groundguard import FactGate

gate = FactGate()
gate.record_tool_result("q3_revenue", "5.2", "billion_usd")
report = gate.check("Q3 revenue came in at $4.8 billion [fact:q3_revenue].", required=["q3_revenue"])
print(report.output_claims[0].status)  # contradicted

python -m pip install groundguard-ai
groundguard-init --template github-action
groundguard-demo
groundguard-benchmark

The PyPI distribution name is groundguard-ai; the Python import name remains groundguard.