The data interface for agents.
1stAuthor defines one interface between agents and deep, domain-specific data: primary sources in each field, parsed to the section, behind the same few verbs. Ask; get the source.
Search was built for people. Agents need something else.
Ten links is not an answer
Web search returns pages for a person to click. An agent gets HTML to parse, ads to skip, and no structure it can rely on.
The deep data is out of reach
Statutes, judgments, full-text papers sit behind PDFs, portals and paywalls. There is no index a program can query with a filter.
All or nothing, cited from memory
Without layers an agent dumps a whole document into context and attributes from recall. IDs get invented; nobody notices.
An Agentic Data Interface fixes the shape: structured, filterable, readable in layers, cited by construction.
Fourteen domains. Each one dug to the bottom.
Also answeringUS law · Trials · Drugs · Grants · Filings · Standards · Vulnerabilities · Federal Register · Organisations · WebEvery domain
One call to a domain. Or let the router choose.
Pick a domain
POST /v1/{domain}/ask — search, read in layers, answer with citations, all inside one call.
Or auto
POST /v1/ask — a router scores every domain for the question and sends it where the sources are.
Web as a safety net
When no domain holds the answer, a general web search steps in — marked as such, never silently.
Those commands are real. While 1stAuthor is in beta the CLI and the Python SDK ship with DeepXiv, who we work closely with: pip install deepxiv-sdk, then deepxiv fa — one token for their papers and for these domains.The deepxiv fa guide
Hand an agent the skill, not this page.
One markdown file per domain, read before the agent calls anything — generated from the same self-description as the docs, so a skill cannot drift from the API.
1stauthorall of it
The three verbs, the credits, the error codes and the working rules — 8 KB, enough to start in any domain.
1stauthor-papersPapers
Search, judge from the card, read one section. Written so an agent stops pulling whole PDFs into context.
1stauthor-casesCases
How to read a judgment in layers, and how to keep every civil claim qualified by coverage.
1stauthor-appraiseAppraise
Resolve a paper by the name people actually use for it, then quote the fellows who cited it.
Start with 2,000 free credits.
No card
Sign in with Google; the account exists the moment you do.
One key, every domain
Scope it, cap it, rotate it from the dashboard.
Pay per read
A search costs 2 credits, a brief 1, a whole document 5, an answer from 50. Failures are refunded and 404s are free.