A short piece of fiction, written by API Evangelist. The reader is imagined; the post, and every number and quotation from it, is real: "Free APIs You Can Call Right Now, Without a Key," Kin Lane, November 13, 2026, https://apievangelist.com/2026/11/13/free-apis-you-can-call-right-now-without-a-key/
Marisol tests software for a living, and on Friday afternoon she was trying to find something for a small AI agent to call. Every API she tried wanted a signup, then an email confirmation, then a card number she didn't want to put in. When she opened the post, its first paragraph described her week: the hardest part is often signing up.
She was suspicious anyway. The post says every API on the list works "without a key, an account or a credit card," and that the author called each example on November 9, 2026, with no key. A claim like that is one she usually checks herself. It also says calling them can't move money, send a message or touch anyone's personal data. That is a lot to promise on someone else's behalf. She decided to trust it a little and verify the rest.
She started with the one the post suggests first. She pasted https://api.frankfurter.dev/v1/latest into a browser tab, and the euro reference rates came back as JSON. It felt almost too easy. She had expected a catch, such as a login wall or a polite error asking her to register.
The catch was in the fine print, and she was glad it was written down. The Bureau of Labor Statistics entry says that without a key you get a smaller daily allowance. The weather entry says the service requires a User-Agent header that identifies you. That explained a failure she'd had a month earlier, when a different service rejected her script. She had blamed the service, and she now suspected her own code.
Then she got a little annoyed, which surprised her. The final section reads like a lecture to her. "Keep request volumes modest." "Cache responses instead of calling every second." She had planned to let her test agent hammer one endpoint in a loop for an hour. She read the advice twice, grudgingly, and saw that the agent would be spending someone's free generosity, not hers. These are mostly run by public institutions and open-source communities, the post says, and she had no good argument against that. Her loop was a convenience for her and a cost for them.
She also wasn't sure the list suited her job. Her agent needed messy, realistic data, and a rhyming dictionary was hardly that. Still, Datamuse was described as having tiny responses that are good for chaining calls, and chaining calls was what she needed to test.
So she did something small. She didn't build the hour-long loop. She wrote a ten-line script that asked https://api.datamuse.com/words?rel_rhy=code&max=3 once and saved the answer to a local file. She gave the request a User-Agent with her team's shared address, in line with the post's advice: "Set a User-Agent that says who you are." Her agent would read the saved file for the rest of the afternoon.
Before leaving she added a line to the top of the script as a reminder: one call, then cache.