What it does
Code-Mode gives an AI agent a single tool that executes TypeScript code with access to a set of registered tools, instead of exposing each tool to the model directly. The code can chain several tool calls in one execution, and the library captures console output and errors. It also offers progressive tool discovery, auto-generated TypeScript interfaces, and runtime interface introspection. Tools can come from MCP servers, HTTP REST APIs, local JSON or YAML files, and command-line tools, and one workflow can mix several of these.
It suits multi-step workflows that would otherwise need many separate tool calls, and work that mixes several tool sources in one run. The publisher also describes a plug-and-play MCP server for MCP-only clients such as Claude Desktop, which runs through npx and loads tool configurations from JSON. For agents with shell access, a utcp command-line interface can write a configuration file and discover tools from an OpenAPI spec or a plain-English description.
Running it involves a Node.js sandbox for the code, with a configurable timeout that defaults to 30 seconds. Direct TypeScript use requires each transport's plugin package to be installed and imported once at startup. The publisher's README also reports benchmark comparisons against traditional tool calling, which this entry does not verify.
Limitations and requirements
Direct TypeScript use requires Node.js and an installed and imported plugin package for each transport. Tools are only accessible through registered UTCP or MCP servers. Some logins, such as OAuth for Notion, are interactive.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Cline MCP Marketplace first seen 13 Nov 2026 · last seen 17 Nov 2026 · its page on the source
1,660
As shown on cline.bot when we last read it