24 free APIs for your next side project

Choose the interaction before choosing the API. Does the user already have a barcode or DOI? Are they checking a changing value, browsing a collection, or asking for a route calculation? Those lead to different small projects, even when every response arrives as structured data.

These 24 free API resources are grouped by that design decision. A first version could be a pantry lookup, a periodically refreshed desk display, an author explorer or a dependency graph. Each group explains what comes back and the conditions that would change the project.

The source documentation and selected public read responses were checked on 19 October 2026. Some resources were assessed through documentation only. These were modest checks, not complete integrations, browser-runtime tests or a reliability comparison.

Start with an identifier the user already has

A lookup tool has a useful boundary: someone supplies one identifier, the service returns a record, and the interface helps them use it. The main design question is what that identifier actually identifies.

A barcode can populate a pantry catalogue. Open Food Facts provides product names, ingredients and nutrition fields, with category and nutrition searches available too. Community records may have missing or incorrect fields, so the catalogue needs room for corrections. Ingredient and allergen data cannot become a safety guarantee.

Public reads need no account or key. The documentation requests an API-usage form and identifying client, with checked limits of 15 product lookups and 10 searches a minute per IP. Cache product records and use exports for larger collections. Database reuse carries ODbL attribution and share-alike conditions; product images have separate CC BY-SA and possible packaging rights.

A DOI can populate a reading list. Crossref returns bibliographic JSON, including titles, authors, publication information and identifiers deposited by members and trusted sources. A DOI importer could let the reader inspect a match before saving the reference.

No signup is required for public REST access. Supply a mailto contact for the polite pool and follow current rate and concurrency headers. Almost all metadata can be used for any purpose, but abstracts may retain copyright. A record about a paper does not grant permission to redistribute the paper itself.

A postcode can locate a local-information page. Postcodes.io returns coordinates and administrative geography, including councils and parliamentary constituencies. Lookups require neither an account nor a key. The result describes a UK postcode, however, rather than a particular property or delivery address. A postcode's representative point can sit across a boundary from an individual address.

Keep that distinction visible in a council finder. Also check the data licences: commercial reuse of the Northern Ireland data requires a separate LPS licence, unlike the Great Britain data's arrangements.

An artist name can lead to stable music identifiers. MusicBrainz supports searching and following relationships among artists, works, recordings and releases. An album-edition explorer can preserve those different entities rather than treating a matching title as a unique album.

Public metadata needs no key, but free hosted access is non-commercial, requires an identifying User-Agent and allows at most one call a second per client. The database has CC0 core data and CC BY-NC-SA 3.0 supplementary data. Core-data permissions do not make commercial hosted-service access free. No audio or cover art comes with the dataset.

Build a panel that changes over time

A changing number needs more context than a label and a refresh button. These sources illustrate different meanings of “current”: a model forecast, an observation, a prediction, a revised event or a reference rate.

The request policies and reuse terms help determine how each panel can work:

• Open-Meteo accepts coordinates without signup or a key and returns current modelled conditions plus hourly and daily forecasts. A personal outdoor-plan or garden dashboard fits the offer. Its hosted free service is non-commercial, including a restriction on ad-supported and subscription projects. Returned data is separately CC BY 4.0, requiring credit. That licence does not remove the service restriction, and modelled conditions are not necessarily measurements at the chosen location.
• TfL supplies status, arrivals, disruptions and journey information for London transport. The current products page allows 50 anonymous requests a minute, or registration for the 500-a-minute product. Follow the transport-data licence and acknowledgements, and account for predictions becoming stale or missing.
• The Environment Agency offers registration-free station, level, flow and warning data, principally for England, under the Open Government Licence with its requested acknowledgement. The documented measurement interval does not guarantee equally prompt delivery. This is explicitly unsuitable for safety-critical reliance: a warning display should link to official warning channels.
• NESO supplies half-hourly estimates and forecasts for Great Britain, excluding Northern Ireland. Its public JSON endpoints need no key, and data is CC BY 4.0 under provider terms. The field called actual is still an estimate. An intensity display does not measure the user's savings or show electricity prices.
• USGS has ready-to-map GeoJSON feeds for rolling windows of recent earthquakes. Use those for a changing display; the catalogue serves date, magnitude and geographic queries. Public feeds need no key. Estimates can change, and USGS-produced data is in the US public domain, with separate rights possible for contributor material. An exploratory map is not a warning system.
• Frankfurter needs no key and supplies daily reference rates, historical values and time series. Default blended rates may be used commercially under the documented terms; institution-specific sources retain provider conditions. A displayed conversion is neither a live tradable price nor the rate a bank will charge.
• GOV.UK Bank Holidays is a small public JSON feed with titles, dates and notes. Choose England and Wales together, Scotland, or Northern Ireland, and refresh the cached calendar. Reuse follows GOV.UK's general OGL conditions. A bank holiday does not establish a particular workplace's, shop's or school's opening hours.
• arXiv's legacy query API returns Atom XML, with titles, authors, summaries, categories and links, rather than JSON. Its terms make descriptive metadata, including abstracts, CC0; paper files retain submission-specific rights. Allow at most one request every three seconds, with one connection at a time across all machines under your control. A scheduled topic digest fits that policy better than querying on every keystroke. Inclusion on arXiv does not establish peer review.

A proposed weekend panel could combine the postcode lookup above with Open-Meteo and the appropriate holiday calendar. Each supplies a different piece: location, forecast and dates. The weather service's non-commercial condition would still apply to the combined project.

Let the user follow a record into a collection

Browsing introduces another question: what can the next click mean? It might move from a work to an edition, from a writer to a related entity, or from a species name to historical observations.

Open Library searches normally return works, with edition information available separately. A personal reading-list tool could use title and author lookup without requiring the reader to type every bibliographic field. Its public APIs are intended for human-facing, mission-aligned use. Identify the application, register its use case as requested and cache results. Bulk work belongs with the monthly dumps; the API is not a high-traffic commercial backend. Catalogue availability supplies neither full-text access nor permission for every cover, and the licensing statement is not a blanket CC0 grant for all contributions.

Wikidata can supply relationships and identifiers across collections. Entity JSON includes labels, descriptions, aliases, statements and links; narrow SPARQL queries can answer relational questions for an author explorer or timeline. Structured data is CC0, while linked media and Wikipedia prose have separate terms. Inspect references and statement ranks, identify the client under the access policy, and use dumps for bulk exports.

Wikipedia's MediaWiki REST API serves a different part of a reading companion: page search and content, with revision and licence information. Language editions can be explored independently; they are not guaranteed translations. Public reads need no key, but redisplayed text generally has CC BY-SA 4.0 attribution and ShareAlike conditions, with exceptions. Preserve revision context, check images individually and follow current client limits.

GBIF connects species-name matching to published biodiversity occurrence records. A species explorer could resolve synonyms, then map records from selected datasets. Public searches require no key; bulk occurrence downloads require an account. Select licences compatible with the project: datasets can be CC0, CC BY or CC BY-NC, and associated media needs separate checks. An observation's date and dataset matter because a record does not establish current presence or absence.

For a different kind of catalogue, TVmaze provides shows, episodes, cast and country-filtered schedules. A cast explorer could connect an actor to other shows, or the same data could support a next-episode calendar. The public API needs no key and explicitly documents browser CORS support. Its data is CC BY-SA, requiring attribution and ShareAlike. Metadata is not streaming permission, and accessible images are not necessarily public-domain images.

Radio Browser exposes station metadata and stream links filtered by language, country or tags. A language-practice directory could browse stations before handing a chosen stream to a player. The free API allows free and non-free software, with server-discovery and rotation expectations rather than a production app pinned to a sample host. Directory-data permissions do not grant broadcast or logo rights. Playback can also fail because a stream is offline, geo-blocked or incompatible.

Ask the service to calculate or match something

Three resources offer more than retrieving a catalogue record. Their results also carry assumptions worth preserving in the interface.

openrouteservice calculates routes, distances, durations and areas reachable within a travel time. A walking-versus-cycling comparison could use route geometry; a reachable-area display could use an isochrone. The €0 Standard plan needs a HeiGIT account and API key, with operation-specific quotas. Results carry CC BY-SA 4.0 and attribution conditions, including openrouteservice/HeiGIT and OpenStreetMap credit. Mapped-network and profile choices affect the result; a computed route is not confirmation of safe or accessible passage.

deps.dev v3 provides package metadata and computed dependency graphs, with graph resolution for npm, Cargo, Maven and PyPI. An explorer could compare two versions or follow the packages bringing in a dependency. Access is free and unauthenticated. The result resembles a clean generic 64-bit Linux resolution, rather than the reader's exact installation, and can contain resolution errors. Generated data is CC BY 4.0, upstream data retains its conditions, and licence fields need independent verification.

OSV matches package/version or commit queries to known vulnerability records, including affected ranges, timestamps and references. Package queries use read-only POST; individual records can be retrieved by GET, with no credentials in the public examples. A dependency advisory explorer could combine those records with the deps.dev graph. Preserve errors and original advisory links: missing results cannot establish safety. OSV aggregates source databases, whose terms matter when redistributing records; the service software's licence is not a blanket returned-data licence.

Treat the asset as another retrieval step

For a gallery or visual quiz, the record and the media are separate ingredients. Artist, title, date and medium can explain an answer, while an image URL or identifier tells the project where to request the picture.

• The Met Collection API returns artwork metadata, image URLs and isPublicDomain, without an API account or key. Its Open Access release covers selected collection data and public-domain images under CC0. A themed art quiz should check the flag and available image fields; a copyrighted work's metadata does not free its image. Use current examples: the docs introduce paginated /v1.1/search and schedule the old search route for retirement on 1 October 2026. Object details remain under /v1/objects/{id}.
• The Art Institute of Chicago API supplies artwork records and image identifiers for separate IIIF URLs. Its documentation explains image sizes and a 60-request-a-minute anonymous limit per IP; no authentication is required. Filter is_public_domain and inspect object rights for a public-domain gallery, because IIIF can also serve copyrighted images. Artwork descriptions are CC BY 4.0; other artwork response data is CC0 under the documented conditions.
• NASA's Image and Video Library API searches descriptions and links to asset manifests for images, video and audio. A mission-history explorer could filter by year, NASA centre and media type. The checked public search needed no key; no numerical quota was located in the inspected guide. Check individual credits and NASA's media guidelines: NASA material is generally not copyright-protected in the US, but third-party content, people and logos can introduce other conditions. This service's access does not describe every NASA API.

Pick the first interaction you can keep small

For a browser-only project, check CORS on the exact endpoint before settling the architecture. No-key access alone does not establish browser access. TVmaze documents it, while some of the other resources here have only documentation or server-side read checks behind them. Client-identification and request policies may also make a cached server-side fetch more suitable.

A pantry lookup can begin with one barcode and editable missing fields. An author explorer can begin with one work, its editions and a source link. A river display can begin with one station, its unit and observation time. Those first screens expose the important data decisions before a project grows around them.

Want to talk about something I’ve written or built? Get in touch.

This article was adapted with AI assistance from an original article on billiem.uk. The original article was reviewed before publication.