What Is Blackridge?

Blackridge is Northwatch Systems' AI runtime economics and forensics platform. It runs as a token economics control point for model traffic and reconstructs AI spend behavior with request-level evidence: which workflow created the spend, which requests were retries or fallbacks, which failed calls still cost money, which agent branches never contributed to the final output, and which traffic has weak or missing attribution.

AI usage is operational infrastructure now. A single product feature may call several models, retry failures, fall back between providers, branch into subtasks, and call tools. The invoice only says money was spent. Blackridge produces the evidence about why — graded, reproducible, and body-free by default (metadata, IDs, timing, usage, pricing, routing, and lineage are recorded; prompts and completions are not).

The investigation UI opens on spend movement, findings, ownership, and evidence quality instead of a provider invoice total.

How it deploys

In customer deployments, Blackridge ships as a small set of binaries and container images. Public self-serve installation is not available; the deployment model is designed around guided assessments and customer rollout:

TrafficExisting AI appOpenAI-compatible or Anthropic-style model traffic.

Control PointToken economics control pointRecords body-free economic evidence as traffic passes through.

EvidenceLedger and reportsUsage, cost, attribution, routing, lineage, and provenance.

InvestigationUI and exportsFindings Finance can inspect and Engineering can reproduce.

Component	What it does
gateway	Deployed inline for OpenAI-compatible and Anthropic-style calls; records canonical economic evidence as traffic passes through
brgctl	Operator CLI for release validation and controlled exports
brg-investigation-api	Read-only GraphQL service over stored evidence; backs the investigation UI, assessment views, and exports

How evidence gets in:

• Runtime mode — point your clients' base URL at the gateway. Highest evidence grades; attribution usually resolves with zero code changes.

The investigation API reads the captured evidence and powers the assessment workflow, request ledger, workflow replay, and evidence exports.

Workflow lineage shows which calls belonged to the run and which calls contributed to the final output.

The ledger keeps the raw economic evidence available for drill-down, replay, and export.

Where to go next

1. Deployment model — see how Blackridge runs in an assessment or customer deployment.

2. Your first assessment — traffic → investigation API → assessment UI.

3. How attribution works — the resolver ladder and why provenance grading matters.

4. How findings are computed — audit the reasoning before you trust the numbers.

Want to see an investigation first? Watch a walkthrough.