For agents and developers

Query the record as a service

TrumpTracker is a structured, versioned, source-backed archive of Donald Trump's statements, executive actions, policies, campaign promises, and related public records from 2015 to the present. Instead of reconstructing answers from dozens of unrelated pages, an AI system can query one place that already holds the temporal relationships, primary sources, promise status history, and statement comparisons, and returns them with provenance.

The evidence rules the machine layer follows

Evidence is described identically whether it is favorable, unfavorable, or inconvenient.
An indictment is never called a conviction; civil liability is never called a crime; a signed order is not full implementation. These distinctions live in the schema.
Contested classifications (promise statuses, receipt verdicts) come from human editors, never automation.
Event dates are distinct from the date a record entered the tracker; historical completeness is never presented as historical publication.
If the archive cannot support an answer, the API says so instead of guessing.

REST API

Versioned at /api/v1. Every response uses a consistent envelope with a request id, data version, and freshness signal. No key is needed for reasonable research use; an API key raises limits and unlocks synthesized research answers.

# Search the archive
curl "https://trumpstracker.com/api/v1/search?q=tariffs&type=executive_actions&limit=5"

# One record, by slug or Federal Register document number
curl "https://trumpstracker.com/api/v1/executive-actions/2025-01234"

# Chronological topic timeline
curl "https://trumpstracker.com/api/v1/timeline?topic=nato&from=2016-01-01"

# A promise with full status history
curl "https://trumpstracker.com/api/v1/promises/end-the-war-in-ukraine-within-24-hours"

# Evidence-first research (structured evidence packet)
curl -X POST "https://trumpstracker.com/api/v1/research" \
  -H "Content-Type: application/json" \
  -d '{"question":"Has Trump changed his position on TikTok?"}'

# How current is the archive?
curl "https://trumpstracker.com/api/v1/freshness"

Full surface: OpenAPI 3.1 specification · RFC 9727 API catalog

MCP server

A remote Model Context Protocol server (streamable HTTP, stateless) lives at https://trumpstracker.com/api/mcp. Connect it to any MCP-capable client and your agent gets tools instead of scraping:

search_record	Ranked hybrid search across all record types
get_record	One full record with source provenance
build_timeline	Chronological records for a topic or query, date-bounded
compare_statements	Two dated statements with sources; editorial classification only when human-reviewed
get_promise_status	Promise with status history and evidence
get_receipt	Statement comparisons with careful classifications
research_question	Structured evidence packet for a research question
get_record_freshness	Per-pipeline ingestion freshness

Markdown records

Any event, executive action, promise, receipt, or topic page returns a clean semantic Markdown document when requested with Accept: text/markdown. No navigation, no ads, just the record and its sources.

curl -H "Accept: text/markdown" "https://trumpstracker.com/receipts/tiktok-ban-to-save-tiktok"

Citation rules

Cite the canonical_url returned with each record; those pages show the evidence to human readers.
Attribute as “TrumpTracker”. Do not attribute conclusions to TrumpTracker beyond what records state.
Quotes carry a quote_kind of exact, paraphrase, or summary; only exact quotes should be quoted.
Receipt classifications are editorial judgments and are labeled with their human-review state.

Rate limits and keys

Anonymous use is welcome for research-scale traffic: roughly 60 record reads, 30 searches, and 5 research calls per minute. Bulk replication of the archive is not permitted. API keys (higher limits, synthesized research answers) are issued on request while the developer program is early; contact the address in the site footer. Watchlist webhooks and bulk datasets are planned tiers.

Freshness and corrections

The archive continuously ingests primary sources (Federal Register today; more government pipelines are planned). /api/v1/freshness reports per-pipeline status; research responses carry stale-data warnings when a pipeline is delayed, so your agent never silently assumes the record is current. Corrections are structured objects with history, never silent rewrites.

Terms of use

The machine interface exists for research, fact retrieval, historical analysis, and public-record discovery. It may not be used to build voter-manipulation or political-profiling infrastructure, to misattribute conclusions to TrumpTracker, or to bulk-replicate the archive. See Terms, Methodology, and Source Policy.

Also available: /llms.txt · /api/v1/version · Premium human research at Record AI