Dicaria Docs

Ask, and get an answer with its sources. Dicaria packs official public data from 10+ countries and regions into one package you call directly: every result carries its reasoning, evidence quotes and a source URL you can open. No naked lists.

Data areas

  • Company data — company master records and disclosures: qualifications, listings, customers & suppliers
  • Economy data — the country-level denominator layer: macro indicators, official statistics, reference rates and the trade mirror
  • ESG data — environmental disclosures and payments to governments
  • Healthcare data — pharma & medical devices on US/EU/AU official sources: approvals, trials, labeling, shortages
  • Legal data — due-diligence red flags and linked disclosures
  • Procurement data — government contract awards: Brazil PNCP, Colombia SECOP II (buyer × supplier)
  • Research data — funded projects, scholarly works and citation impact (NSF / NIH / OpenAlex …)
  • Security data — open-source dependency vulnerabilities (OSV) and the CISA KEV catalog

Quick start

agentdata is one pip package with two query families: corpus queries take a single access token (PAT) and reach every collection; realtime queries run on your machine, direct to upstream public sources with no key.

Corpus queries Realtime queries
What you get Pre-built company profiles and lists, continuously collected from official public disclosures Live lookups: identifier resolution, list screening, disclosure search
Where it runs Hosted API (access-token PAT authentication) Your machine, direct to upstream public APIs (no account, no key)
Rate limits Per-key quota No platform-side quota; each upstream's own limits apply
Caching — Local automatic cache (SQLite): within its lifetime the same query is served from cache, without hitting upstream
pip install agentdata

# See everything you can ask first (capability discovery, zero config) — one line per face: summary, jurisdiction, execution plane, parameter closed sets
agentdata capabilities --format md

# 30-second try (realtime, zero config) — a company name → GLEIF LEI candidates
agentdata company gleif_lei_records_search --filter-entity-legal-name "Siemens" --page-size 3

# The remaining examples omit the leading executable name — prepend `agentdata ` (or `python -m agentdata `) for a full command.

# Clinical trials a company is recruiting for (realtime; source: ClinicalTrials.gov)
healthcare ctgov_studies --filter.advanced 'AREA[LeadSponsorName]("AstraZeneca") AND AREA[OverallStatus](RECRUITING)' --page-size 3

# List search (corpus, once access is opened) — every row carries evidence and sources
company search --filter company_registration_country=US --limit 5
company read --id 1476 --level tier2

Corpus queries need an access token — self-serve: sign up in the Console and create an API key; the free allowance arrives with your account.

export AGENTDATA_BASE_URL=https://dicaria.com/api/corpus
export AGENTDATA_API_KEY=mk_live_…   # create it under Console → API keys; bound to an allowed data-area list

Connect via MCP

Wire Dicaria into your agent (Claude Code, Cursor, Windsurf…): the MCP endpoint authenticates with the same access token (PAT) — configured independently of the SDK's AGENTDATA_API_KEY.

claude mcp add --transport http --header "Authorization: Bearer <PAT>" dicaria https://mcp.dicaria.com/agentdata/mcp
  • Endpoint: https://mcp.dicaria.com/agentdata/mcp (transport: http; streamable)
  • Auth: Authorization: Bearer mk_live_… — replace <PAT> with a real key created under Console → API keys; no anonymous mode
  • Or just tell your agent "Install https://dicaria.com/": it reads the home page and configures itself as above

Where to start