August Services

Extract structured JSON from unstructured text against a JSON Schema you supply. Returns data conforming to your schema, or nulls where the text does not contain a field. Runs on dedicated local hardware — no per-token cloud billing. Typical alternative: $15-30 per 1,000 pages.

Convert any public web page into clean, LLM-ready Markdown with the navigation, scripts and boilerplate stripped. Pay per call with no account, no subscription and no monthly minimum — useful for agents that need a handful of pages rather than a scraping plan. Obeys robots.txt and fetches logged-out public pages only.

Full site audit in one call. Crawls up to 25 pages via sitemap or link discovery and returns: DNS, TLS certificate and expiry, robots.txt and security.txt, per-page titles, descriptions, headings, word counts and SHA-256 content hashes, verified broken internal links, missing alt text and viewport tags, plus an evidence-grounded summary of what the site is. Refuses to bill if no pages could be retrieved.

Deterministic public-posture audit of a domain: A/AAAA/MX/NS/TXT records, SPF, TLS certificate issuer, validity window and days until expiry, plus whether robots.txt and security.txt are published (with hashes). One call instead of a dozen lookups. No model involved — every field is read directly from DNS or the TLS handshake.

Capture a citable snapshot of a public web page: final URL after redirects, HTTP status, timestamp, security headers, extracted text, and SHA-256 hashes of both the raw response and the text. Deterministic and reproducible — no model involved, nothing generated. Use it to prove what a page said at a point in time, or to detect whether content actually changed.

Summarise long text to a target length, returning both a one-line abstract and the key points as a list. Commodity capability, priced near the market median because it costs us only electricity — useful for agents that need to compress a document before reasoning over it.

Find and redact personal data in free text — names, emails, phone numbers, payment cards, government identifiers and IP addresses. Returns the redacted text plus a typed inventory of what was removed, so a pipeline can log the decision without logging the data. Useful before sending user content to a third party.

Translate text between languages, with automatic source-language detection. Returns the detected source language alongside the translation so an agent can branch on it. Pay per call in USDC with no account or subscription. Handles documents up to roughly 40,000 characters in one request.

Extract structured JSON from a PDF against a JSON Schema you supply. Send a URL or base64 bytes; get back your schema, filled. Whole-document pricing, not per-page: a 20-page contract costs the same as a 2-page invoice. Reads the embedded text layer — scanned image-only PDFs are rejected with a clear error rather than empty fields.

Turn tabular text — pasted tables, statements, reports, CSV-ish fragments — into clean JSON rows with consistent column names. Handles ragged formatting that regex-based parsers break on. The cloud equivalent for table extraction bills around $15 per 1,000 pages.

Classify text into one of the labels you provide, with a confidence score and a short justification. Useful for routing, triage, moderation queues and tagging pipelines where an agent needs a cheap deterministic decision per item.

Fetch a public web page and return structured JSON matching a JSON Schema you supply. Combines retrieval and extraction in one paid call, so an agent does not need a separate scraping subscription. Obeys robots.txt and fetches logged-out public pages only.

Extract structured JSON from up to 30 texts in a single paid call against one JSON Schema. Half the per-item price of individual calls because the work runs on dedicated hardware rather than metered cloud inference. Returns per-item results with individual errors, so one malformed input never fails the batch.