Pi

Minimal harness by badlogic, now under the earendil-works org. Best-in-class design clarity; measured 7.2M tokens on our suite at 6/10 resolved.

Pi — specifications
GitHub stars109K (updated September 2026)
LanguageTypeScript
LicenseMIT
Bring your own keyYes
PricingFree, open source; bring your own provider
Footprintn/a
Sandboxn/a
Tokens on our 10-task suite7.2M (resolved 6/10)

Repository: earendil-works/pi

On our benchmark suite

10-task SWE-bench Verified subset, five agents through the same LiteLLM proxy on Z.ai GLM-5.3, same provider, 600 s per-task cap, vanilla configs. Pi measured 7.2M tokens, resolving 6/10 of tasks. Methodology and per-task data: full report.