timeout

basicwidespread

Timeout is a preset waiting limit: if a program or service does not respond within this time, the system stops waiting and signals an error.

In plain terms

Imagine calling customer service and deciding to hold the line for at most five minutes. If no one answers by then, you hang up and try again later or seek another solution. Timeout is the same rule in the world of computers.

Programs constantly wait on each other: your browser waits for the website, the banking app for the bank's server, the AI agent for the model or an external tool. If the other party freezes or the connection breaks, the waiting program would hang forever without a timeout.

Timeout is therefore not the problem itself, but a protective limit. It indicates that something was too slow or got stuck, and gives the system the opportunity to retry, choose a different path, or provide a clear error message to you.

How it works

The developer or operator sets a deadline, for example 30 seconds, for a response to a request.
When the program sends the request, an internal timer starts.
If the response arrives within the deadline, the stopwatch stops, and everything continues as usual.
If the deadline expires, the program interrupts the wait and reports a timeout error.
Subsequently, depending on its configuration, the system retries the operation, switches to a backup solution, or displays an error message to the user.

An everyday example

You are paying at an online store; the loading indicator spins for a long time, then a message appears: 'The request timed out, please try again.' The same happens when an AI chatbot, after long thinking, returns an error message instead of an answer, or when a coding agent stops a command that has been running for too long.

Who is affected

everyoneOn websites, apps, and AI chatbots, anyone encounters 'time out' error messages, and it is good to know these are usually temporary.

developerIt must decide how much waiting is reasonable and what happens when the time expires. → Code

bank customer, investorDuring payment or transfer, after a timeout, the transaction may not have failed, so it is worth checking the account before retrying. → Money

entrepreneur, executivePoorly configured timeouts cause stuck orders, unfinished agent tasks and dissatisfied customers.

A common misconception

MythIf I get a timeout, the operation definitely did not happen.

In realityTimeout simply means the response did not arrive in time. The request may still have reached its destination and been executed, so check the result before retrying.

Limits and risks

A timeout does not reveal the cause of the error: it could be a slow network, an overloaded server, or a frozen program.
A timeout that is too short needlessly interrupts slow but working operations, while one that is too long keeps you waiting and ties up resources.
Automatic retry can cause duplication, such as a double-placed order or a double-sent message, if the system does not recognize that the same request has arrived again.
With AI agents, long, multi-step tasks can easily time out, leaving work in an unfinished state.

Similar, but not the same

latency – Latency measures how long it takes to receive a response, while timeout is the limit after which we no longer wait for it.
deadline – The deadline is a fixed point by which the entire task must be completed, while the timeout sets the allowed duration for a single wait.
rate limit – The request limit determines how many requests you can send within a given time, while the timeout determines how long you wait for a single request's response.

Related concepts

Broader

error handling, reliability

Narrower

connection timeout, session expiry, gateway timeout

On the edge

retry, latency, rate limit, API, tool calling

A brief history

Timeout is one of the oldest protective mechanisms in computer networks. The 1981 description of the TCP protocol (RFC 793) already required the sender to retransmit data if it receives no acknowledgment within a given time. Methods were later worked out for adapting this waiting time to the network's measured speed (RFC 6298, 2011). RFC 793 was replaced by RFC 9293 in 2022.

Where things stand

It is part of practically all networking software. AI agents give it new attention because the usual deadlines of a few seconds are often too short for tasks that run for minutes or hours. Such tasks are therefore increasingly run in the background, and the program periodically asks how far along the work is.

Status: 2026 November

Also known as

timeout · timeout · time-out · time out · time limit exceeded · wait time expired · request timeout · gateway timeout

Sources

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.