This paper describes an attempt to provide a linear algebra package suitable for small microcomputers. Those linear algebra facilities which are implemented are a set of the most popular currently used on mainframes; that is, the solution of linear systems, linear least squares, and the symmetric eigenvalue problem. Memory is at a premium on small microcomputers, and so to provide all these facilities simultaneously an SVD algorithm is implemented as the underlying code for each requirement, exploiting features of BASIC to keep the code compact. By these means, the user’s own time and the microcomputer’s memory are used economically, whereas the (usually) less important microcomputer’s actual computing time is lengthened to some extent, in comparison with the most efficient possible codes for each requirement. A nonlinear least squares algorithm, based on the repeated use of the linear least squares code, is also implemented.