# AI Tools for Developers

A Docker-based tool, now deprecated in favor of Docker Desktop's MCP Toolkit, that ran agentic workflows written in Markdown prompt files against dockerized tools and any LLM.

AI Tools for Developers ran agentic AI workflows built from Markdown prompt files and dockerized tools, letting a user pick their own LLM. Tools were packaged as Docker images so an agent could take more complex actions, work across environments, and run in a sandboxed container; each prompt could be configured to use a different LLM model or family, so a frontier model could plan while lightweight local models executed steps.

A conversation loop passed tool results, agent responses, and prompt content back and forth, retrying with different parameters or tools when an agent hit an error. A serve mode let registered Markdown prompts and their tools be used as MCP servers, referenced by git reference or local path. A project-first design extracted context about the current project from a dedicated extractor image.

The repository is deprecated and no longer under active development; the README directs users to Docker Desktop's integrated MCP Toolkit instead. It is written in Clojure under the Apache-2.0 license.

Website: https://www.docker.com/blog/introducing-docker-mcp-catalog-and-toolkit/