Terminal Tool Platform

Toolipie — Code Snippet Toolbox

A local-first Python CLI/TUI platform that turns small one-off scripts into indexed, reusable tools with manifests, logs, and plugin boundaries.

Toolipie is less a gallery of snippets and more a small operating surface for personal utilities. The live repository frames it as a Python CLI platform for centralizing small one-off scripts so they stop being rewritten or forgotten.

That changes the article shape: the interesting part is the registry, manifest, CLI/TUI execution model, and the way tiny tools become dependable enough to reuse.

01

The Tool Registry Is the Center

Toolipie is built around discovery first: scan tools, write an index, then let CLI and TUI read the same registry.

toolipie — registry workflow
$ toolipie scan
 refreshed .toolipie/index.json

$ toolipie list
md-to-docx      Markdown to Word documents
md-to-pdf       Markdown to styled PDFs
pdf-to-png      Render PDFs into image folders
png-prep-ocr    Prepare PNGs for OCR pipelines
md-to-docx coremd-to-pdf corepdf-to-png corepng-prep-ocr corepng-to-avif pluginpng-to-svg plugin
Why this matters

The registry means the interface does not need to import every tool just to show what exists. Discovery stays fast, and code can load lazily only when a run starts.

02

A Manifest Before a Tool

The core abstraction is not a folder of Python files. It is a declared interface that says what the tool is, what it needs, and how it should run.

Toolipie separates small utilities into core tools and plugin tools, but both use the same manifest shape. A tool declares a key, title, summary, default glob, options, and requirements. The index tracks whether it came from core or plugin space.

schema_version: 1
name: md-to-pdf
title: Markdown to PDF
summary: Render Markdown files into styled PDFs
default_glob: "*.md"
options:
  - name: preset
    type: string
    default: a4_report
Core tools

Stable, lightweight utilities that live under the maintained platform tree.

Plugin tools

Installable or packaged utilities that can be added without rewriting the platform.

Safety boundary

Manifests are validated and plugin code is not imported during discovery.

03

Execution Feels Like a Terminal App

The TUI is not decoration. It gives small scripts a repeatable run surface with options, output, progress, and cancellation.

The repository describes a two-pane terminal UI: navigate tools, inspect descriptions, open options, run jobs, and watch output in an in-TUI console. Runs happen in a subprocess under a PTY, which keeps progress rendering consistent and gives cancellation a predictable path.

  • Options from manifests: parameters are generated from each tool's schema plus common input/output fields.
  • Run logs: individual tool runs can write structured logs such as `run.jsonl`.
  • Packaging flow: plugins can be packaged into zip files and installed back into the repo.
  • Cancel model: TUI runs can request termination and escalate if a process does not exit.
run panel
$ toolipie run pdf-to-png \
  --input input/pdf-to-png \
  --output output/pdf-to-png \
  --param dpi=300
 output/pdf-to-png/report/page-001.png
04

From Utility Shelf to AI-Callable Tools

The long-term direction is not more random scripts. It is a local tool platform that can be discovered by humans and, eventually, AI workflows.

Current platform

Core tools, plugin folders, a registry, a CLI, and a TUI make local utilities easier to find and run.

Near-term polish

Better scaffolding, installer flows, and docs would make adding a new personal utility less manual.

AI layer later

The repository's vision includes AI assistance for tool creation: scaffold code, configure options, generate docs, and potentially expose tools through MCP-like workflows.

Product judgment

Toolipie succeeds if a useful local action stops being a forgotten file and becomes a durable command.