Quickstart

Goal

Get from a real agent workflow to a reviewable go/no-go decision. The runner executes locally and uploads normalized metadata to the hosted control plane.

What should happen

The scan should produce a run record, normalized findings, an evidence pack reference, and a decision a reviewer can approve, waive, or block.

Create a runner token

Create a scoped runner token in settings. Store it in CI or a local shell as ASSURANCE_INGEST_TOKEN.

local environment

export ASSURANCE_API_URL="https://app.policystrata.com"
export ASSURANCE_INGEST_TOKEN="runner_token_..."

Select the project

Create or pick the project slug for the agent workflow under review.

project identity

export ASSURANCE_PROJECT="support-bi-copilot"
export ASSURANCE_RELEASE="release_2026_07_06"

Run local checks

Run DocPull and PolicyStrata where the customer docs, traces, schemas, and policies already live.

docpull sync \
  --source customer-support-docs \
  --out .assurance/docpull

Upload evidence

Upload normalized run metadata and artifact references.

upload evidence

agent-assurance evidence upload \
  --project "$ASSURANCE_PROJECT" \
  --release "$ASSURANCE_RELEASE" \
  --docpull .assurance/docpull \
  --policystrata .assurance/policystrata

Review the release

Open the review console, inspect blockers, attach a decision, and export the evidence pack for audit storage.

Expected Outputs

Output	Purpose
`assurance-run.json`	Normalized run metadata, tool versions, inputs, status, and findings.
`evidence-pack.json`	Reviewable release bundle with summary, run refs, artifacts, and decision state.
Review console run	Reviewer-facing history, blockers, waivers, and approvals.

Runner

Configure runner inputs, exit codes, uploads, and local artifact storage.

Release gates

Decide which findings block CI and which require review.