Skip to content

Bananas

jaunt is a new way of coding by using python embedded specs.

Code Demo Cerebral Valley

Core trick

Developers write Python signatures and docstrings, and agents generate the function bodies from those embedded specs.

Tags

dev toolsagents

Description

jaunt is a new way of coding by using python embedded specs. we move the implementation logic to the agents, while business logic is still written and maintained by the devs.
ijindal1/jaunt★ 00 forksCreated Jul 5, 2026 · last push Jul 5, 2026

README excerpt

# Jaunt

<div align="center">
  <img src="docs-site/public/images/tyger-blake-1794.jpg" alt="William Blake, 'The Tyger' from Songs of Experience (1794). The Metropolitan Museum of Art, Open Access." width="280" />
  <br/>
  <sub>William Blake, <em>The Tyger</em>, plate 42 from <em>Songs of Experience</em> (1794).
  <a href="https://www.metmuseum.org/art/collection/search/347983">The Metropolitan Museum of Art, Open Access.</a></sub>
</div>

<br/>

> *Tyger Tyger, burning bright,*
> *In the forests of the night;*
> *What immortal hand or eye,*
> *Could frame thy fearful symmetry?*
>
> -- William Blake, via Alfred Bester's *The Stars My Destination*

Jaunt is a small Python library + CLI for **spec-driven code generation**. You
write the intent — a signature and a docstring — and Jaunt writes the
implementation under `__generated__/` using the OpenAI Codex CLI (`codex exec`).

Call `jaunt.magic_module(__name__)` once at the top of a file and every top-level
stub below it becomes a spec, with no per-symbol decorators:

```python
import re
import jaunt

jaunt.magic_module(__name__, prompt="All parsers are RFC 5322 strict.")

EMAIL_RE = re.compile(r"...")           # real body → handwritten, kept as-is

class Email:
    """Email with from_, to, subject, body. Validates on construction."""
    # docstring-only class → Jaunt designs and writes the whole class

def parse_email(raw: str) -> Email:
    """Parse an RFC 5322 payload into an Email. Raise ValueError on malformed
    input,

Team