Metadata-Version: 2.4
Name: seamspec
Version: 0.0.1
Summary: Executable specs for AI coding agents. Build first; every seam that works gets a spec.
Keywords: ai,coding-agents,specs,testing,mcp,claude-code,regression
Author: vamship
License-Expression: Apache-2.0
Classifier: Development Status :: 1 - Planning
Classifier: Environment :: Console
Classifier: Intended Audience :: Developers
Classifier: Programming Language :: Python :: 3
Classifier: Topic :: Software Development :: Quality Assurance
Classifier: Topic :: Software Development :: Testing
Requires-Python: >=3.11
Project-URL: Homepage, https://github.com/seamspec/seamspec
Project-URL: Repository, https://github.com/seamspec/seamspec
Description-Content-Type: text/markdown

# SeamSpec

**Executable specs for AI coding agents. Build first; every seam that works gets a spec.**

> **Status:** pre-release (0.0.1). Under active development.

SeamSpec is an open-source governance layer for AI coding agents such as Claude Code, Cursor, and VS Code Copilot.

- When you and your agent get something working, SeamSpec helps turn that behavior into a **small, human-approved test** (an *invariant*).
- Approved invariants are **enforced automatically**: by editor hooks while the agent works, by a git pre-commit check, and in CI.
- Failed approaches are remembered as **dead ends**, so the agent doesn't retry them.

Most spec tools ask you to write the spec *before* you build. SeamSpec writes it *after*, from code that works, and makes sure it never silently breaks.

The name comes from Michael Feathers' *seams*: places where untested code can be put under test without rewriting it.

## Install (pre-release)

```bash
uv tool install seamspec
seamspec
```

## Links

- Repository: https://github.com/seamspec/seamspec

## License

Apache-2.0
