# ScreenVerity > Per-call U.S. exclusion/debarment and licence screening for software and agents. One HTTP POST > replaces a portal browse-and-parse session, and every answer comes with an Ed25519-signed receipt > naming the exact snapshot of each list that was checked. Screening one name through a state or federal portal usually costs an agent a browser session, HTML parsing, retries and pagination — on the order of 50k-200k tokens (estimate), with a real chance of partial extraction and no durable audit artefact. One `POST /v1/screen` is typically ~2k tokens of request and response, covers every list currently loaded, and returns a receipt you can verify later. This is compliance-workflow tooling, not a consumer report. A `possible_match` is a prompt to adjudicate, never an automatic rejection. ## What it is not Coverage is exactly what `GET /v1/sources` reports and nothing more — `full`, `partial` or `never_loaded`, per source, with the age of each snapshot. There is no claim of 50-state completeness. If a list could not be checked, the response says `complete: false` and names it. Lists are never silently skipped. ## Endpoints - [Agent skill doc](https://api.screenverity.com/SKILL.md): tools, request shape, and when to call. Start here. - [OpenAPI 3.1](https://api.screenverity.com/openapi.json): machine-readable spec for every route. - [Sources](https://api.screenverity.com/v1/sources): free. Every loaded list, jurisdiction, coverage, snapshot age, record count. - [Health](https://api.screenverity.com/health): free. Liveness, source count, active record count. - `POST /v1/screen`: $0.25 USDC via x402. Unpaid calls return HTTP 402 with payment requirements; pay and retry. - `GET /v1/receipts/pubkey` and `POST /v1/receipts/verify`: free. Check any receipt's signature and hashes. ## MCP An MCP server exposes three tools — `screen`, `sources`, `verify_receipt` — over stdio, so an agent can call this instead of reaching for a browser. Without a funded wallet, `screen` returns an explicit `payment_required` error and states that no lists were checked; it never returns a fabricated `clear`. ## Optional - [Receipt format and matching model](https://api.screenverity.com/openapi.json): confidence bands, decisive rules, and the `norm_v`/`model_v` stamps that identify which matcher produced a result.