This paper extends the Mirrlees (1971) model of optimal non-linear income taxation with a monitoring technology that allows the government to verify labor effort at a positive, but non-infinite cost. Monitored individuals receive a penalty, which increases if individuals earn a lower income (provide less work effort) or have a higher earning ability. We analyze the joint determination of the non-linear monitoring and tax schedules and the conditions under which these can be implemented. Monitoring of labor effort reduces the distortions created by income taxation and raises optimal marginal tax rates, possibly above 100 percent. The optimal intensity of monitoring increases with the marginal tax rate and the labor-supply elasticity. Our simulations demonstrate that monitoring strongly alleviates the trade-off between equity and efficiency as welfare gains of monitoring are around 1.4 percent of total output. The optimal intensity of monitoring follows a U-shaped pattern, similar to that of optimal marginal tax rates. Our paper explains why large welfare states optimally rely on work-dependent tax credits, active labor-market policies, benefit sanctions and work bonuses in welfare programs to redistribute income efficiently.