Toolshop

The Tool Framework for AI Agents

Usage

from toolshop import all_tools
from marvin.beta import Application

app = Application(
    tools=all_tools()
)

app.say("Write mergesort.py")
app.say("Test mergesort.py")

Description

Toolshop is a general framework for developing tools used by LLM Agents. By developing tools with Toolshop, your tools inherit many helpful features.

• Configuration: Toolshop tools are classes, not functions. This enables users to configure their tools for a particular session before passing them to the agent.
• Confirmations: Toolshop makes it easy to add required user confirmations to tool calls. Examples might include an agent editing sensitive files or provisioning cloud resources.
• Constraints: Tools can share state. This allows users to enforce constraints that require more context than the tool call arguments alone. Examples might include prevention of multiple parallel edits to the same file, requiring file reads before file edits, and cost-based limits on tool usage.
• Compatibility: Toolshop aims to be compatible with most major LLM frameworks. Avoid framework lock-in and easily use the same tools with different frameworks.

Design Philosphy

Design for AGI

Agents will continue to improve [ref]. Developers should spend time building things that benefit from improved agents rather than suffer. Tool development is a great example of work that directly benefits from improved agents. The recent movement of AI companies away from simple text-completion models and towards tool-assisted chat completion models reinforces the importance of tool development [ref, ref].

Constrain tools, not process

Prompts that correct agent behavior by heavily constraining process may actually hold back more intelligent agents that do not require such corrective hand-holding. By instead enforcing constraints in the tools themselves, the planning and reasoning process native to the underlying agent is unhindered. The reasoning and planning capabilities of agentic AI offered by companies such as OpenAI and Anthropic will continue to improve over time.