A confidence interval is a random interval in a parameter space with a defined coverage probability of the true parameter value. The coverage probability is called "confidence". From a sample of observed data (realizations of random variables with some given distribution), an observed interval can be constructed, which is a realization of the random interval. It is important to understand that the coverage probability is a feature of the random interval (i.e. of the procedure of how the interval is to be constructed based on data), not of the observed interval.

There are different ways and strategies to construct confidence intervals. The are often constructed by inverting a hypothesis test about the parameter (the interval includes all values that would not be rejected in a hypothesis test).

To understand these statements you should read an introductory stats book and read some articles about confidence intervals that you can easily find in the internet.

It comes with uncertainty management. If you look for the body weight of a population X (mean = M), you sample a subpopulation x and you find a mean weight in this population (m_x). However your estimation depends on the subpopulation x that you sampled.

Then you obtain confidence intervals around your m_x estimate (I don't go to the maths here). In frequentist statistics the 95% CI mean that if you repeat your sampling many times 95% of the confidence intervals area contains the true value of interest m.

It is quite counterintuive to interpret and many people would interpret this 95%CI as a bayesian 95% credible intervals which simply means that you are 95% that the true value of interest M is in your confidence interval area.