Tempa

Tempa — Agentic Coding Automation Harness

[](https://github.com/daus95/tempa/actions/workflows/tests.yml) [](https://github.com/daus95/tempa/releases) [](LICENSE)

Turn a spec into a working app — clarified, planned, implemented, and QA'd automatically, while you do something else.

Turning a spec into working software usually means endless manual back-and-forth: clarifying vague requirements, breaking the work into a plan, implementing it feature by feature, and catching bugs before they pile up — one agent session at a time, babysat from start to finish.

**Tempa automates that entire loop.** It drives an agentic coding CLI — **Claude Code**, **GitHub Copilot CLI**, or **OpenAI Codex CLI**, your pick per stage — repeatedly and unattended: clarifying your PRD until nothing critical is left ambiguous, drafting a plan (epic → feature → task), implementing it, and running QA — start it once, walk away, and come back to real progress instead of a blank terminal. It runs on whichever CLI's login you already have (Claude, GitHub Copilot, or ChatGPT/OpenAI) — no separate per-token API billing to provision or watch. See [Choosing a CLI Backend](#choosing-a-cli-backend) below.

Official repo: [github.com/daus95/tempa](https://github.com/daus95/tempa)

Once set up, the recommended way to drive Tempa is the web **dashboard** (`tempa dashboard`) — see [Dashboard (Recommended)](#dashboard-recommended) below. Every dashboard action also has an equivalent CLI command (`tempa `), for power users or scripting.

---

Contents

1. Quick Start
2. Setup (One-Time Only)
3. [Choosing