Anthropic AI has launched a beta of its “tool use” functionality for all Anthropic Message API users for the Claude suite of generative artificial intelligence (AI) models.
The beta allows users to connect Claude to external tools, which enables real-time information lookup, access and integration of third-party features into Claude’s workflow. Essentially, this means users can insert simple code snippets into their account API interface that give Claude the ability to bring outside information into its conversations.
One example given by Anthropic involves the use of about 10 lines of code to integrate a third-party stock ticker into the model.
With this integration, the model would no longer rely on web lookup to discover stock prices — a paradigm that can result in the model confidently displaying incorrect data — but instead can pass along information verified at the source level to be correct.
Because tool use for Claude uses a chain-of-thought reasoning model, the AI system is able to process complex commands requiring multiple third-party access integrations. This allows Claude to answer simultaneous queries and process requests that require step-by-step information seeking and integration.
According to Anthropic, all of the available Claude AI models can handle numerous tool requests:
“All models can handle correcting [sic] choosing a tool from 250+ tools provided the user query contains all necessary parameters for the intended tool with >90% accuracy. These limits apply to the total number of tools, regardless of complexity.”
The crypto, Web3 and blockchain communities could potentially benefit from Claude’s capabilities as, for the most part, any integration developed for the platform should be usable from within the API structure.
The use cases for crypto verticals could range from integrating real-time, trusted price tickers all the way to full decentralized exchange interface interoperability. Users could even connect wallet managers and portfolio trackers to the model and create trading workflows.
It bears mention that AI models can make mistakes and that Claude is a general chatbot not specialized for financial, cryptocurrency, or blockchain uses specifically. This means even the most trustworthy integrations could still end up producing incorrect outputs.
Related: Google eyes paid AI search features, explores subscription model