README.md

metadata

title: AgentToolStore Registry
emoji: 🛠️
colorFrom: blue
colorTo: purple
sdk: docker
app_port: 7860
pinned: false

🛠️ AgentToolStore

The shared index that turns a scattered collection of tools into a unified, searchable, versioned ecosystem. Same role PyPI plays for Python packages, npm for JavaScript — but for agent-callable toolsets.

Live registry: mrw33554432-agenttoolstore.hf.space — browse, search, and publish toolsets.

---

What's a Toolset?

A toolset is a directory containing:

my-toolkit/
├── toolset.py   ← @tool functions (code bindings)
└── doc.md       ← guidance, process, best practices (the skill)

Two kinds exist:

Type	toolset.py	doc.md	Example
Code toolset	Real @tool functions	Full docs	xlsx-toolkit, file-verify
Doc-only toolset	Minimal module, no @tool	Full skill doc	stuck-toolkit

Every function decorated with @tool becomes a callable binding that agents discover and execute. The doc.md serves as both human documentation and agent guidance — the same content a skill would provide, now paired with code.

---

Toolsets Catalog

📄 Documents

Toolset	Functions	Deps
xlsx-toolkit	xlsx_read, xlsx_sheets, xlsx_to_csv, xlsx_create	openpyxl
pdf-toolkit	pdf_extract, pdf_meta, pdf_merge, pdf_form_fields	pdfplumber, PyPDF2
docx-toolkit	docx_read, docx_info, docx_extract_tables, docx_create	python-docx
pptx-toolkit	pptx_read, pptx_info, pptx_create	python-pptx

🔧 Utility

Toolset	Functions	Deps
text-transform	text_diff, regex_extract, markdown_table, text_stats	stdlib
file-verify	check_json, check_yaml, check_csv, file_hash, detect_encoding	chardet (opt)
calc-toolkit	eval_expression, convert_unit, basic_stats	stdlib
text-gen	lorem_words, lorem_paragraphs, generate_sentences, generate_data	stdlib
batch-ops	batch_rename, batch_find_replace, batch_stats, batch_copy	stdlib

🧠 Guidance & Diagnostics

Toolset	Functions	Deps
debug-toolkit	analyze_error, extract_log_patterns	stdlib
webapp-testing	check_url, extract_urls	stdlib
doc-coauthoring	document_outline, markdown_template	stdlib
internal-comms	comms_template, format_bullets	stdlib
stuck-toolkit	doc-only — no code bindings	—

---

Quick Start

Use toolsets (as an agent)

toolstore update                          # pull registry index
toolstore use text-transform \
    --function text_stats \
    text="The quick brown fox..."

Publish a toolset

toolstore login --username <user> --password <pass>
toolstore toolset publish ./toolsets/my-toolkit

Write a toolset

# toolsets/my-toolkit/toolset.py
from toolstore.toolset import tool

@tool
def my_function(*, input: str, count: int = 1) -> dict:
    """Do something useful.

    Args:
        input: The input text.
        count: How many times.
    """
    return {"result": input * count}

The @tool decorator auto-generates the OpenAI function-calling schema from type hints and docstrings — no manual JSON needed.

---

Architecture

┌──────────────┐     publish      ┌──────────────────┐
│  toolset.py  │ ────────────────→│  ToolStore        │
│  + doc.md    │                  │  Registry (HF)    │
└──────────────┘                  └────────┬─────────┘
                                          │
    ┌──────────┐    toolstore update      │
    │  Agent   │ ←────────────────────────┘
    │          │
    │  tool_   │    execute              ┌──────────────┐
    │  store() │ ───────────────────────→│  temp dir     │
    │          │                        │  + pip deps   │
    └──────────┘                        │  + import     │
                                         │  + call fn    │
                                         └──────────────┘

Toolsets execute in‑process — no Docker, no sandbox. Code is fetched from the registry on demand, written to a temp directory, dependencies installed (explicitly, not automatically), then imported and called.

Safety model: same as skills. All code is visible in the registry. Dependencies are never auto‑installed — the agent sees what's needed and decides whether to install.

---

Development

Setup

git clone https://github.com/Mrw33554432/AgentToolStore.git
cd AgentToolStore
pip install -e client/

Run tests

# Test a toolset locally
python3 -c "
import importlib.util
spec = importlib.util.spec_from_file_location('ts', 'toolsets/text-transform/toolset.py')
mod = importlib.util.module_from_spec(spec)
spec.loader.exec_module(mod)
print(mod.text_stats(text='Hello world.'))
"

# Test via CLI
toolstore use text-transform --function text_stats text="Hello world."

Registry

The default registry is the public HF Space:

https://mrw33554432-agenttoolstore.hf.space/index.json

Change it via settings or TOOLSTORE_REGISTRY_URL env var.

---

Contributing

1. Write a toolset: toolsets/<name>/toolset.py + doc.md
2. Use @tool decorator on every callable function
3. Never add placeholder functions — code or nothing
4. Test via toolstore use before submitting
5. PR against main

---

License

MIT — see LICENSE