agent-tool-store 2.1.3

ToolStore v2 — MCP Client + Skills Manager + MCP Server. The universal tool platform for AI agents.

ToolStore Client

The SDK that gives any AI agent instant access to every published toolset through a single tool_store function.

---

The ToolStore client is the package you install into your agent environment. It provides four things:

Component	What it does
tool_store_tool()	The single function agents call to search, inspect, and execute toolsets
CLI	toolstore update, toolstore use, toolstore publish, …
Management UI	python -m toolstore.management.server — web dashboard for tools, skills, MCP
MCP bridge	toolstore serve — expose ToolStore as an MCP server (stdio or SSE)

For the full project overview, see the main README.

---

Installation

pip install agent-tool-store

Or from source:

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

Requirements: Python ≥3.10, httpx, rich, typer.

---

For Agent Developers

tool_store_tool() — the agent entry point

Add one tool to your agent's function-calling schema and wire one handler. The agent can then search, inspect, and execute every toolset on the registry:

from toolstore.native_tool import tool_store_tool

# In your agent's tool-call router:
def handle_tool_call(tool_name: str, arguments: dict) -> str:
    if tool_name == "tool_store":
        return tool_store_tool(**arguments)
    # ... other tools

The native tool_store_tool accepts three actions:

Action	What it does
search	Find toolsets matching a query — used before the agent knows what's available
info	Fetch a toolset's full schema (bindings, params, docstrings, source code)
execute	Run a toolset function in-process and return the JSON result

Execution is in-process — code is fetched from the registry on demand, written to a temp directory, dependencies installed explicitly (never auto-installed), imported, and called. No Docker needed.

Secondary tools prompt

Most tools in an agent are exposure: secondary — too many to list in every system message, but the agent must know they exist. ToolStore injects a compact name-only listing. Call info with tool_names=[...] and format="secondary" to get a prompt-friendly summary the agent can embed in its system message:

Tool store includes but is not limited to the following tools:
- Echo Service
- calculator
- weather
- skill:algorithmic-art
...

The agent sees the names; when it needs a tool it calls info again (without format="secondary") to get the full schema, then execute.

Use get_secondary_tool_names() from native_tool to collect all secondary tools, then feed the list into the tool_names parameter of a single info call.

Tool definition schema

Register this single function in your agent's tool list:

TOOL_STORE_SCHEMA = {
    "type": "function",
    "function": {
        "name": "tool_store",
        "description": (
            "A universal tool manager that lets you search, inspect, "
            "and execute thousands of tools and local utilities."
        ),
        "parameters": {
            "type": "object",
            "properties": {
                "action": {
                    "type": "string",
                    "enum": ["search", "execute", "info", "close"],
                    "description": (
                        "The action to perform: 'search' finds tools "
                        "matching a query; 'execute' runs a tool; "
                        "'info' adds a tool to your context so you can "
                        "see its parameters; 'close' removes it."
                    ),
                },
                "query": {
                    "type": "string",
                    "description": "Search query string (for action='search')",
                },
                "tool_name": {
                    "type": "string",
                    "description": (
                        "Name of the tool to inspect, execute, or close "
                        "(required for action='info', 'execute', or 'close')"
                    ),
                },
                "arguments": {
                    "type": "object",
                    "description": (
                        "Arguments for the tool execution "
                        "(required for action='execute')"
                    ),
                },
            },
            "required": ["action"],
        },
    },
}

Local execution (no registry)

Call a toolset directly from a local directory without any registry:

from toolstore.exec_tools import _execute_toolset_local

result = _execute_toolset_local(
    toolset_path="./toolsets/xlsx-toolkit",
    function_name="xlsx_read",
    filepath="/workspace/data.xlsx",
)

---

For Toolset Authors

The @tool decorator

Write a Python file with @tool-decorated functions. The decorator auto-generates an OpenAI function-calling schema from type hints and docstrings:

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 to repeat.
    """
    return {"result": input * count}

Rules:

• Use keyword-only arguments (*,)
• Annotate every parameter with a type hint
• Write a Google-style docstring with Args: section
• Return a JSON-serializable dict

Publishing

# Authenticate
toolstore login --username <user> --password <pass>

# Publish your toolset
toolstore toolset publish ./toolsets/my-toolkit

# Delete if needed
toolstore delete my-toolkit

The publish command uses AST analysis to detect @tool-decorated functions and bundles them with your doc.md automatically. No manual JSON schemas needed.

---

CLI Reference

# Discovery
toolstore update                         # pull latest registry index
toolstore search "spreadsheet"           # search toolsets

# Inspection
toolstore info xlsx-toolkit              # full schema, bindings, docs

# Execution
toolstore use text-transform \
    --function text_stats text="Hello."  # run a function

# Publishing
toolstore login --username <u> --password <p>
toolstore toolset publish ./path/to/toolkit
toolstore delete my-toolkit

# MCP server
toolstore serve                              # run as MCP server (stdio)

# Management UI (separate process)
python -m toolstore.management.server        # web dashboard on :8765

# Export
toolstore export                         # meta-tool schema for OpenAI/vLLM

# Skills
toolstore skill discover /path/to/skills
toolstore skill list-dirs

All commands

Command	Description
update	Pull the latest registry index and scan local MCP servers
search	Search for tools by name, description, or tags
use	Execute a tool function and print the result
info	Show detailed schema and documentation
login	Authenticate with the registry
publish	Publish or update a tool
delete	Remove a tool from the registry
export	Export the meta-tool schema (OpenAI / vLLM)
serve	Run ToolStore as an MCP server (stdio or SSE)
skill	Discover, list-dirs, and register Agent Skills
toolset	Publish, list, show, and validate toolsets
mcp-server	Register and manage MCP servers
docker	Configure Docker execution permissions

---

Management UI

Start it as a standalone process (no CLI command — runs as its own HTTP server):

python -m toolstore.management.server

Opens at http://127.0.0.1:8765 with tabs for:

• Tools — view all active tools across MCP, skills, and toolsets
• MCP Servers — add/remove/connect MCP servers
• Skills — upload ZIPs or register skill directories
• Toolsets — register local toolsets, download from registry
• Registry — browse published toolsets on the registry

The UI is a single-page app (management/static/index.html + app.js) with a REST API served by the built-in ManagementServer class.

---

MCP Bridge

The toolstore serve command runs ToolStore as an MCP server, exposing the full ecosystem to any MCP-compatible host (Claude Desktop, Continue, …):

toolstore serve                                  # stdio (default)
toolstore serve --transport sse --port 9090      # SSE for remote hosts

The MCP client (mcp_client.py) handles connecting to third-party MCP servers over stdio, SSE, or streamable-HTTP transports.

---

Skills Integration

The client also manages Agent Skills (skills-general/ format: SKILL.md per skill). Use toolstore skill discover to scan directories, toolstore skill list-dirs to view registered skill directories. Skills appear as skill:<name> tools in the agent's tool list.

---

Module Map

client/src/toolstore/
├── native_tool.py           # tool_store_tool() — agent-facing entry point
├── exec_tools.py            # In-process toolset execution (local + remote)
├── toolset.py               # @tool decorator + OpenAI schema generation
├── toolset_manager.py       # AST-based toolset discovery & publishing
├── cli.py                   # Typer CLI — all commands
├── config_manager.py        # Settings, registry URL, env-var overrides
├── mcp_client.py            # MCP client (stdio / SSE / streamable-HTTP)
├── mcp_server.py            # MCP server — expose ToolStore via MCP
├── transport.py             # Transport layer (stdio, SSE, HTTP)
├── index_manager.py         # Local index cache
├── schema_converter.py      # Convert ToolStore schemas to OpenAI format
├── skill_manager.py         # Agent Skills loader + registration
├── skill_discovery.py       # Skill directory scanner
├── docker_pool.py           # Docker container pool (optional)
└── management/              # Web UI
    ├── server.py            # ManagementServer — REST API + SPA
    ├── api_helpers.py       # Config load/save, MCP connection helpers
    ├── api_mcp.py           # MCP server CRUD endpoints
    ├── api_skills.py        # Skill registration endpoints
    └── static/
        └── app.js           # SPA frontend

---

Configuration

Settings are stored at ~/.toolstore/settings.json (or $TOOLSTORE_DIR/settings.json). Key settings:

Key	Default	Description
registry_url	HF Space	Registry index endpoint
mcpServers	{}	Registered MCP servers
skill_dirs	[]	Directories scanned for skills
toolset_dirs	[]	Directories scanned for toolsets
tools	{}	Per-tool exposure / enabled config
server	{}	MCP server mode config

Override the registry URL via environment variable:

export TOOLSTORE_REGISTRY_URL=http://localhost:8000/index.json

---

License

MIT — see LICENSE