They’re small. Really small. From the Congressional Research Service (Jan. 5, 2015):

Because job projections, in particular, involve numerous assumptions and estimates, the State Department’s job estimates for Keystone XL have been a source of disagreement. One challenge to State’s analysis is that different definitions (e.g., for temporary jobs) and interpretations can lead to different numerical estimates and “fundamental confusion” about the Final EIS numbers. Consequently, it may be difficult to determine what overall economic and employment impacts may ultimately be attributable to the Keystone XL pipeline or to the various alternative transport scenarios if the pipeline is not constructed.

The CRS report quotes this portion of the State Department EIS:

During construction, proposed Project spending would support approximately 42,100 jobs (direct, indirect, and induced), and approximately $2 billion in earnings throughout the United States…. Construction of the proposed Project would contribute approximately $3.4 billion (or 0.02 percent) to the U.S. gross domestic product (GDP). The proposed Project would generate approximately 50 jobs during operations. Property tax revenue during operations would be substantial for many counties, with an increase of 10 percent or more in 17 of the 27 counties with proposed Project facilities. …

Additional clarification on what proportion of those jobs would be American, vs. foreign, sourced is from this Cornell report:

Based on jobs information provided by TransCanada for the FEIS, KXL US on-site construction and inspection creates only 5,060-9,250 person-years of employment (1 person-year = 1 person working full time for 1 year). This is equivalent to 2,500-4,650 jobs per year over two years. (page 7)

What about the materials used in the construction of the pipeline?

TransCanada claims that “the $7 billion (KXL) pipeline project is expected to directly create more than 20,000 high-wage manufacturing jobs and construction jobs in 2011-2012 across the Us, stimulating significant additional economic activity.”20 This claim is misleading and erroneous on a number of levels.
First, as discussed above, the budget for KXL US that relates to incremental US employment is $3 to $4 billion and not the $7 billion claimed by the proponents. Second, TransCanada and other KXL proponents are giving the impression that KXL will create a high number of manufacturing jobs. This is simply not true. The main manufacturing activity related to pipeline construction is the manufacture of the steel pipe. The 36-inch steel pipe is the largest single materials input for KXL. This is literally the pipe in the pipeline. In general, pipeline construction is not a manufacturing-intensive activity even if the steel itself is also being manufactured onshore.
This section will present strong evidence that:
(a) almost half (and perhaps more) of the primary material input for KXL—steel pipe—will not even be produced in the United States;
(b) based on the experience of Phases 1 and 2, the final processing work for KXL will probably be performed in the US with most of the steel and pipe sourced from oustide of the US (notably India and South Korea).21 (page 11)

Now, no doubt there will be some spillover effects (Note: If you didn’t believe in spillover/multiplier effects arising from the American Recovery and Reinvestment Act (ARRA), you won’t believe in any multiplier effects from Keystone XL construction.). Let’s assume there is such a thing as a spending multiplier; then using plausible estimates for incremental building, and accounting for imported inputs, the Cornell researchers conclude:

…the incremental US spending associated with KXL project construction is only about $3 to $4 billion. Given a multiplier of 11 person-years per $1 million, this translates into total employment impacts of 33,000 to 44,000 person-years.59 So a reasonable estimate of the
total incremental US jobs from KXL construction is about one-third of the figure estimated in the Perryman study and used by industry to advocate for the construction of KXL.
Moreover, any job impacts associated with KXL construction would be spread over 2 and more likely 3 years.60 So the annual impacts are at most about 22,000 person-years of employment per year, for two years.61 But the annual impacts could also be as low as 11,000 person-years per year, for three years.62 (page 24)

For context, the US economy added 252,000 to nonfarm payroll employment in December alone.