1 of 10

Using command-line tools to power content strategy

2 of 10

A little bit about me

​

  • Senior Devops Engineer @ Kiteworks
  • Google Developers Expert (Cloud)
  • Book Author for "Golang for Jobseekers"�https://www.amazon.com/Golang-Jobseekers-Unleash-programming-advancement-ebook/dp/B0C7ZVS44M

​

3 of 10

Managing a blog is a pain

4 of 10

Ideas for what content to add

  • Look through past content to see what new content can be combined together (let the AI read)
  • Look through current coding projects (let the AI read)
    • Get the rough gist of the new coding project
    • Expand on an interesting problem within the coding project
  • Look through website analytics to find out what the blog is well known for and build on that (how to get AI to access it)

5 of 10

Connecting an AI to external sources

  • MCP CLIs / Servers
  • Plain CLI (Bash commands)

6 of 10

Connecting models to MCP for getting data

(Advantages)

  • Centralized approach for handling it
    • Good if needed on org level - easier to keep everyone's mcp server updated at one go
  • Industry standard

7 of 10

Connecting models to MCP for getting data

(Disadvantages)

​

  • Usually, all tools are pre-loaded on start
  • Context heavy
    • Common complaint - mcp responses tend to be verbose
  • MCP is somewhat complex to wrap around
    • No basic curl / wget etc to try things with it
  • Easy to deploy but getting it usable for centralized approach is hard
    • Authentication / Authorization

8 of 10

Just use CLI

  • Let the model use the bash shell
  • It knows how to navigate - just leave it be
  • Certain default behaviours
    • Non zero exit if command is a failure
    • Help flags available for commands and main commands
    • Really good cli will also provide assistance of what flags etc could be missing etc

9 of 10

Example

https://github.com/hairizuanbinnoorazman/Go_Programming/tree/master/Apps/content-tools

10 of 10

Thanks