How Feed America Operates — Seven Values

1. Open data, CC BY 4.0

Every dataset Feed America publishes is released under Creative Commons BY 4.0. That means anyone — researchers, journalists, AI companies, partner aggregators, even competitors — can use, share, adapt, and build on the data, including for commercial purposes. Attribution is the only requirement.

We don't believe public food-help data should live behind paywalls. The federal sources we ingest (USDA FNS, HRSA, NCES) are public domain. The community pantry data we aggregate via Plentiful, ampleharvest, Salvation Army, and partner imports has always been meant to be shared.

Surfaces: /research, /atlas.json, /hsds/v3/

2. Free forever, no API keys, no rate limits for reads

No registration, no login wall, no per-call billing, no rate limits for read access. Anyone with curl can pull our entire dataset. AI agents querying us get the same access as enterprise partners.

The only friction we add is fair-use rate limits on writeable endpoints (5 submissions/hour per IP at /submit) to prevent spam. Read access is unrestricted forever.

Surfaces: /api/openapi.json (no auth required), /feeds

3. No gating: never charge a pantry to be listed

Every food pantry, food bank, soup kitchen, mobile pantry, community fridge, WIC clinic, FQHC, and SNAP-authorized retailer is in the directory for free. We never charge for a listing. We never charge for a "verified" badge. We never charge for status updates. We never charge for hours edits.

Pantries that claim their listings on /pantry get the same treatment as pantries we discovered via Plentiful or USDA — including the free "Verified on Feed America" badge they can embed on their own website.

4. Brand disambiguation: protect donor intent

Feed America (EIN 92-1761881, Houston) is a separate, unaffiliated 501(c)(3) from the larger homonym charity Feeding America (EIN 36-3673599, Chicago). The names are similar; the entities are independent. Donor confusion has cost both organizations real money over the years.

We treat donor disambiguation as a first-class operational responsibility. Every page references the EIN. Every donor-facing surface explicitly disambiguates. Our disclosures page walks through the homonym in plain language. Always verify the exact EIN on IRS.gov before donating.

Surfaces: /disclosures, /about, /donate.json

5. Infrastructure-first: ~98% of dollars to platform + data

Feed America operates on a lean cost structure. We have no staff payroll. ~98% of donor dollars go to platform infrastructure (Cloudflare Workers, D1 database, Pages CDN), data verification, partner coordination, and open-data publication. Founder Sharika Parkes operates the org in a volunteer capacity.

This is a deliberate choice. Tech-leveraged charities at our stage compound better with infrastructure than with payroll. As we grow, we'll add paid staff for verification ops, partner onboarding, and translation — but only with explicit donor consent and full disclosure on Form 990.

Surfaces: /donate/where-your-gift-goes, /transparency, /api/v1/health.json (live operational stats)

6. Spanish parity, not afterthought translation

~13% of the US population speaks Spanish at home. Spanish-speaking households face disproportionate food insecurity. We invest in full Spanish parity for every public surface — not as an afterthought translation, but as a first-class language.

That means Spanish gets its own LLM-search detection, its own atlas + glossary + calendar + per-state disaster pages, its own hreflang chains, its own Schema.org inLanguage='es-US' tagging. The same data freshness rules; the same operator workflows.

Surfaces: /es, /es/atlas, /es/glossary, /es/calendar, /api/llm-search (bilingual)

7. AI-first discoverability

More people ask AI assistants for food help every quarter. Feed America's data needs to be the canonical source AI agents cite — accurately, in real-time, with proper attribution.

We invest aggressively in AI discoverability: Schema.org JSON-LD on every page, dedicated machine-readable JSON endpoints (atlas.json, snapshot.json, donate.json, about.json, etc.), MCP server with 7 tools for direct AI client integration, OpenAPI 3.0 spec with 83+ endpoints, robots.txt explicitly allowing 14 AI crawlers, llms.txt + llms-full.txt discovery files, a master /api/v1/ai-tools.json integration manifest.

When an AI agent answers "where can I find food in Houston" or "how do I apply for SNAP in Texas", the answer should cite Feed America with structured data — not hallucinated.

Surfaces: /api/v1/ai-tools.json, /.well-known/mcp.json, /llms.txt