Estela reconstructs billable hours from Claude Code and Codex transcripts plus git history, so developers who skip timesheets recover the time they worked.
Yes, Estela is open source under the MIT license.
Estela has a free tier plus paid plans. The open-source version can be self-hosted with no license cost; you pay only for your own infrastructure.
Yes, Estela can be self-hosted (the source is available under the MIT license).
Estela reconstructs billable hours from Claude Code and Codex session transcripts plus git history, aimed at the developers and agencies who bill hourly but never actually fill in a timesheet. Open source: yes (MIT); self-hostable, with a free tier that runs locally and Pro/Teams tiers that add hosted sync on an open-core model. It is freemium and, honestly, brand-new — 0 stars at the time of listing — and is maintained by jdleonruiz.
Estela solves a specific, common problem: the work happened, but nobody wrote down the hours. It reads the transcripts of Claude Code and Codex sessions together with git commit history and infers how much billable time was actually spent, so the record is reconstructed from artifacts you already produce rather than from memory at the end of the week. The free tier runs entirely locally as a CLI; the paid Pro/Teams tiers add hosted sync for sharing across a team — a standard open-core split. Open source: yes (MIT); self-hostable; freemium.
For an agency or a fractional operator, unbilled time is lost revenue, and reconstructing it after the fact from AI-session transcripts and commit history is a genuinely useful angle — the evidence of the work already exists, so Estela just turns it into an hours record. The honest read: this is a day-one project with 0 stars and no maintenance history, so nothing about its reliability or continuity is established, and inferred hours from AI sessions and commits are an estimate — review them before they become an invoice. Watch it as an interesting idea rather than a dependency until it accumulates a track record.