T cells are key elements of cancer immunotherapy1, but certain fundamental properties, such as development and migration of T cells within tumours, remain elusive. The enormous T cell receptor (TCR) repertoire, which is required for recognising foreign and self-antigens2, could serve as lineage tags to track these T cells in tumours3. Here, we obtained transcriptomes of 11,138 single T cells from 12 patients with colorectal cancer (CRC) and developed STARTRAC (single T-cell analysis by RNA-seq and TCR tracking) indices to quantitatively analyse dynamic relationships among 20 identified T cell subsets with distinct functions and clonalities. While both CD8+ effector and “exhausted” T cells exhibited high clonal expansion, they were independently connected with tumour-resident CD8+ effector memory cells, implicating a TCR-based fate decision. Of the CD4+ T cells, the majority of tumour-infiltrating Tregs showed clonal exclusivity, whereas certain Treg clones were developmentally linked to multiple TH clones. Notably, we identified two IFNG+ TH1-like clusters in tumours, the GZMK+ TEM and CXCL13+BHLHE40+ TH1-like clusters, which were associated with distinct IFN-γ-regulating transcription factors, EOMES/RUNX3 and BHLHE40, respectively. Only CXCL13+BHLHE40+ TH1-like cells were preferentially enriched in tumours of microsatellite-instable (MSI) patients, which might explain their favourable responses to immune-checkpoint blockade. Furthermore, IGFLR1 was highly expressed in both CXCL13+BHLHE40+ TH1-like and CD8+ exhausted T cells and possessed co-stimulatory functions. Our integrated STARTRAC analyses provide a powerful avenue to comprehensively dissect the T cell properties in CRC, which could provide new insights into the dynamic relationships of T cells in other cancers.