billiem

Home /

24 free APIs for your next side project

24 API resources for weather, local data, books, art, music and software, with project ideas and current access conditions.

  • Choose APIs by what your project should show or do.
  • Free-service conditions, data licences and media rights can differ.
  • Proposed combinations connect local information, reading lists and dependency graphs.

A postcode can become a local dashboard. A barcode can fill in a pantry catalogue. A museum collection can supply the paintings for an art quiz. Free APIs give small projects something real to work with: places, forecasts, books, music, public records and the connections between them.

The useful starting question is what you want the project to do. These 24 resources cover quite different jobs, from returning one bank-holiday calendar to calculating the area you can reach on foot.

Provider documentation and selected public read responses were checked on 19 October 2026. Some entries are based on documentation only. The checks establish a dated starting point, not a reliability comparison or a set of completed integrations. Free access also comes in different forms, so the conditions that change a useful choice are included beside it.

Give a place its own dashboard

Postcodes.io turns a UK postcode into coordinates and administrative information such as a council or parliamentary constituency. That makes it a useful first step for a local-information page: the reader supplies a postcode, and the project has somewhere to centre its results. Public lookups need no account or key.

It returns postcode geography, rather than a property address. A postcode's representative point can fall on a different side of a boundary from an individual address, so keep a council finder appropriately qualified. Its data licences also distinguish Great Britain from Northern Ireland: commercial reuse of the Northern Ireland data requires a separate LPS licence.

Once you have coordinates, Open-Meteo can add current modelled conditions and hourly or daily forecasts. An outdoor-plan page could put rain and wind side by side through the afternoon; a garden dashboard could use the documented soil and weather variables. Responses include times and units, which belong beside the displayed values. “Current” here means modelled conditions, not necessarily a measurement at the reader's location.

The hosted free service needs no key or signup, but it is for non-commercial use. An ad-supported or subscription project counts as commercial. Returned data has a separate CC BY 4.0 licence, requiring credit; that data licence does not remove the hosted-service restriction.

For a London dashboard, the TfL Unified API supplies line status, arrival predictions, disruptions and journey information. A favourite-line card or bus-stop departures display gives it a small, clear job. TfL's current API products page allows anonymous access at 50 requests a minute, with registration for the 500-a-minute product. Follow the transport-data licence and acknowledgements, and show when predictions were refreshed.

openrouteservice adds a different capability: calculation. It can return route geometry, distance and duration, or an isochrone, the area reachable within a chosen travel time. That opens up projects such as comparing a walking route with a cycling route, or showing what lies within a short journey of a selected point.

Its €0 Standard plan requires a HeiGIT account and API key, with separate quotas for directions, reachable areas and other operations. Results carry CC BY-SA 4.0 and attribution conditions, including credit to openrouteservice/HeiGIT and OpenStreetMap. Routing quality depends on the mapped network and selected profile; a calculated route does not establish accessibility or safe passage.

Follow changes in the wider world

Four more sources give a display or map a subject that changes over time:

  • Electricity: the NESO Carbon Intensity API returns half-hourly estimates and forecasts, an intensity category, generation mix and regional information for Great Britain. A desk display could compare the current interval with upcoming lower-intensity periods. It needs no key, and the data is CC BY 4.0 under the provider's terms. Even the field named actual is an estimate. It is not electricity pricing, measured savings or a Northern Ireland feed.
  • Earthquakes: USGS GeoJSON feeds return recent events with magnitude, place, time and map coordinates. Choose a rolling feed for a live display, or the catalogue API for a date, magnitude or geographic query. Public feeds need no key. Event estimates can be revised, so retain their update information. USGS-produced data is in the US public domain; contributor material can have separate rights. A map of recent events is an exploration project, not an earthquake warning system.
  • Rivers: the Environment Agency flood-monitoring API exposes stations, levels, flows and flood warnings, principally for England. A local river card could show the latest value, its units and observation time. Access needs no registration, and reuse is under the Open Government Licence with the agency's requested acknowledgement. Measurement intervals do not guarantee equally prompt delivery. The provider explicitly rules out safety-critical reliance, so a warning display should link people to official warning channels.
  • Wildlife records: GBIF supports species-name matching and searches across published biodiversity occurrence records. A species explorer could resolve synonyms before mapping records from selected datasets. Public searches are available without a key; bulk occurrence downloads need an account. Dataset licences vary between CC0, CC BY and CC BY-NC, so select compatible records and preserve their sources. A record says where something was recorded, not where it is definitely present today. Associated photographs need their own licence check.

Those distinctions suggest what the interface should show. A river card needs a unit and observation time. An earthquake marker needs an event link and update time. A biodiversity map needs the record's date and dataset, rather than a claim that an animal is there now.

Make books and knowledge easier to browse

Open Library's Search API can fill a reading list with titles, authors, publication details and identifiers. Search normally returns works, while a work can have many editions. That is a useful distinction if your project needs to separate a novel from a particular translation or printing.

The public APIs are intended for human-facing, mission-aligned lookup. Identify the application, register its use case as requested, and cache results. Use the monthly data dumps for bulk work; the hosted API is not offered as a high-traffic commercial backend. Catalogue access is neither full-text access nor permission to reuse every cover, and Open Library's licensing statement does not describe all contributions as blanket CC0 material.

A reading tool can go beyond isolated book records with Wikidata. Its entity JSON contains labels, descriptions, aliases, statements and links to other collections. An author explorer could connect a writer to works and external library identifiers; a narrowly scoped SPARQL query could find relationships for a timeline.

Wikidata's structured data is CC0. Linked pictures and Wikipedia prose have their own terms. Inspect statement references and ranks before presenting a claim as settled fact, and choose an entity lookup, query service or dump to suit the size of the job. These public services require client identification and considerate use.

For article text, Wikipedia's project-local MediaWiki REST API provides search and routes for page content, including revision and licence information. A contextual reading companion could offer a source page for an unfamiliar subject. A language-learning browser could compare the available language editions, without assuming they are translations of one another.

Public reads need no key. When redisplaying prose, preserve the attribution, licence and revision context: text is generally CC BY-SA 4.0, with exceptions. Images require a separate file-level check. Wikimedia's current request limits also distinguish identified clients from unidentified traffic, so do not build around old claims of unrestricted access.

Turn research metadata into a reading tool

A DOI is enough to start with Crossref. Its REST API returns bibliographic JSON deposited by members and trusted sources: titles, authors, publication details, identifiers and other fields. That suits a DOI-to-reading-list importer, or a bibliography search box where someone can check a result before saving it.

Public access needs no signup. Supply a mailto contact to use the polite pool, and follow the current rate and concurrency headers. Almost all metadata can be used for any purpose, but abstracts may retain copyright. Neither a DOI nor a returned licence link gives your project blanket permission to republish the paper.

arXiv's legacy API is more directly suited to a topic feed. Query by category, author or keywords and receive Atom XML, with titles, authors, summaries, categories and links. A daily physics or computing digest could collect the results and send readers to the abstract pages. Inclusion on arXiv does not establish peer review.

The API terms make descriptive metadata, including abstracts, CC0. Paper files retain their own submission-specific rights. Keep requests slow: at most one every three seconds, with one connection at a time across all machines under your control. This is a good fit for a periodically refreshed feed rather than a request for every keystroke in a search box.

Build a gallery with something worth looking at

An art quiz needs more than an image URL. Artist, date, medium and a link to the original collection record can make the answer worth reading too.

The Metropolitan Museum of Art Collection API returns those fields alongside image URLs and isPublicDomain. Its Open Access collection releases selected collection data and public-domain images under CC0, making an artwork-of-the-day page or themed guessing game plausible without arranging an API account. Check the public-domain flag and available image fields; metadata about a copyrighted work does not free its image.

Use the current documentation when copying search examples. The Met introduced paginated /v1.1/search, with retirement of the old /v1/search scheduled for 1 October 2026. Object details still use /v1/objects/{id}.

The Art Institute of Chicago API offers another route through artists, periods and media. Artwork records contain an image identifier used to construct a separate IIIF image URL; the documentation explains the supported sizes. No authentication is required, and the documented anonymous limit is 60 requests a minute per IP.

For a public-domain gallery, filter is_public_domain and inspect each object's rights. IIIF can also serve copyrighted images. Artwork descriptions are CC BY 4.0 while other artwork response data is CC0 under the documented conditions, so even fields in the same response do not all have identical terms.

For a different kind of visual explorer, the NASA Image and Video Library API searches descriptions and returns links to asset manifests for images, video and audio. A mission-history page could narrow results by year, media type or NASA centre, then retain the original descriptions and credits. This service uses images-api.nasa.gov; it is not a promise about every API behind NASA's separate developer gateway.

The checked public search needed no key. Before displaying an asset, follow NASA's media guidelines and inspect its credit. NASA material is generally not copyright-protected in the US, but third-party content, identifiable people and logos can introduce other conditions. No numerical quota was found in the inspected API guide, which is not the same as unlimited access.

Catalogue food, television and music

Open Food Facts can turn a barcode into a product name, ingredients and nutrition fields. That is a substantial shortcut for a pantry catalogue: scanning identifies the product, and the user can fill in missing or incorrect information. Category and nutrition searches could support a comparison page as well.

Reads need no account or key, but the docs request an API-usage form and an identifying client. The checked limits were 15 product lookups and 10 searches a minute per IP, so cache records and use exports for larger collections. Follow the database's ODbL attribution and share-alike conditions; product images have separate CC BY-SA and possible packaging rights. Community records can be incomplete or wrong, so ingredient and allergen fields should not become a safety guarantee.

For a watchlist, TVmaze returns show details, episodes, cast and country-filtered schedules. A next-episode calendar has a useful shape before you add recommendations or social features. The public API needs no key, and its docs explicitly describe CORS support for browser requests. Data is CC BY-SA: link back to TVmaze and follow ShareAlike. Show metadata does not provide streaming rights, and image access should not be mistaken for public-domain imagery.

Music offers two quite different projects:

  • Radio Browser supplies station names, languages, countries, tags and stream links. A language-practice directory could let someone choose a language and browse stations instead of choosing a single broadcaster in advance. The free API supports free and non-free software, but applications should discover and rotate available servers rather than hard-code a sample host. Directory data permissions do not grant rights to the broadcasts or station logos; streams can also be offline, geo-blocked or incompatible with the player.
  • MusicBrainz supplies identities and relationships: artists, recordings, works, releases and their identifiers. That suits an album-edition explorer or a personal collection helper, rather than a music player. Public metadata needs no key, but the hosted API is free for non-commercial use, requires an identifying User-Agent and allows at most one call a second per client. Its core data is CC0 and supplementary data CC BY-NC-SA 3.0. Core-data reuse rights do not make commercial use of the hosted service free, and the dataset supplies neither audio nor cover art.

Add a small useful feature

Some APIs are useful precisely because the job is narrow.

The GOV.UK Bank Holidays feed contains event titles, dates and notes for England and Wales together, Scotland, and Northern Ireland. A countdown or project planner can let the reader choose the appropriate division. It is public JSON without an account, reusable under GOV.UK's general OGL conditions. Refresh your cached copy and link the official calendar. Bank holidays alone do not tell you whether a particular shop, workplace or school is closed.

Frankfurter returns currency reference rates, including a date, base currency, quote currency and rate, with historical and time-series options. A travel-budget comparison could show how the same amount converted on different dates. The public API needs no key, and the default blended rates can be used commercially under the documented terms. They are daily reference calculations, not live tradable prices or a quote from someone's bank. Show the observation date, and check provider terms before switching to an institution-specific source.

Explore the software behind a project

deps.dev's v3 API returns package metadata and computed dependency graphs. A tree explorer could show which packages pull in a dependency, or compare the graphs for two versions. Graph resolution currently covers npm, Cargo, Maven and PyPI, while metadata has wider ecosystem coverage. Public access is free and unauthenticated.

The graph represents the service's resolution assumptions, roughly a clean generic 64-bit Linux installation, rather than your exact installed environment. It can include resolution errors. Generated data is CC BY 4.0, source data retains its conditions, and licence metadata still needs independent verification.

OSV complements that graph with known vulnerability records. Queries can use a package and version or a commit; records include affected ranges, timestamps and references. A release comparison could collect relevant advisories and link readers to the original reports. Public examples need no credentials. Package queries use a read-only POST, while individual records can be retrieved by GET.

Keep the output specific: “these known records were returned”, rather than “this version is safe”. OSV aggregates multiple source databases, and missing records or imports remain possible. Check those sources' terms if redistributing their records; the OSV software licence does not establish a blanket licence for all returned advisory data.

A few combinations to try

The interfaces between these services suggest projects with a manageable first version:

  • A personal weekend panel: Postcodes.io supplies a location, Open-Meteo supplies the forecast, and the GOV.UK feed marks the appropriate bank holidays. Start with one saved postcode and a few clearly dated values. Keep the weather service's non-commercial condition in mind if the project grows into an ad-supported site.
  • A reading-list explorer: Open Library finds a work and its editions; Wikidata can add author relationships and external identifiers; Wikipedia can offer a linked background page. Keep catalogue records, structured claims and article prose distinct, with their respective source and licence information.
  • A dependency advisory explorer: deps.dev supplies the computed graph, then OSV can look up known records for packages and versions in it. Show graph-resolution errors and advisory sources rather than turning the result into a security verdict.

For a browser-only version, check the exact endpoint's CORS support before committing to the design. A keyless URL is not enough: TVmaze explicitly documents browser access, while some other services above have only documentation or server-side read checks behind them. A cached server-side fetch may suit a provider's identification and request policies better.

Once the project has a useful subject, Free Compute covers places to run it. The free-assets guide covers icons, fonts, art and sound for the interface. A first version can be as small as one postcode, one book or one river station, with a source link and enough context to make the returned value mean something.