Skip to main content

Accessing portal data with AI coding agents through skills

  • March 16, 2026
  • 2 replies
  • 14 views
S

I’ve been investigating options for accessing the data on our portal using AI assistants, mainly Claude code. I previously posted my experience of a REST to MCP adapter, which works, but is a bit clunky and not very token efficient. I have recently experimented with a new library called mcp2cli which enables the coding agent to use CLI commands to query data on our portal.

MCP2CLI can create a skill which introspects the swagger.json  - which is published under the portal’s API tab, to understand the explore API of the portal. This enables the coding agent to discover datasets, understand their structure and create SQL queries to return filtered and aggregated data. The skill is an alternative approach to MCP servers that is more token efficient due to progressive disclosure and the use of a command line interface to query the data. I iterated over creation of the skill several times to improve it.

My limited testing of the skill I created indicates that it is quite accurate and token - efficient. It also works on GitHub copilot as well as claude code, and I presume opencode, codex etc.

There’s a more in - depth description of this on my blog.

 

    2 replies

    Benwa
    Huwise Team
    Forum|alt.badge.img
    • Benwa
    • Huwise Team
    • March 31, 2026

    Hey Steve, great to see you're still pushing the boundaries here! The mcp2cli approach is genuinely clever — using progressive disclosure via CLI to reduce token overhead is a smart trade-off, and the fact that it works across Claude Code, GitHub Copilot and others makes it quite versatile.

    The token efficiency angle you raise is something we think about too on our end. It's one of the design considerations behind our MCP server — curious whether you've had a chance to compare the two approaches head-to-head in terms of token consumption on similar queries?

    For anyone landing on this thread who missed our earlier discussion on the topic 👉 https://community.huwise.com/the-world-of-data-74/mcp-tools-to-access-portal-datasets-and-catalog-691?postid=1493#post1493

      S

      Hi Benwa

      I haven’t done much more evaluation of the different approaches, partly due to other things taking precedence and also the fact that I only have a Claude Pro subscription and the usage limits are pretty tight at the moment. But it would be good to do a proper evaluation. I focussed mainly on getting the skill working well, and my (unvalidated) perception is that the skill was more robust than the previous MCP approach I tried - I emphasise that this not Huwise’s MCP, it is one I adapted from the rest-to-mcp-adapter library.

      It looks like the CLI paradigm for AI tools interacting with data is gaining momentum when compared to MCP - but it’s such a fast changing landscape, who knows where we will end up!

        Terms & ConditionsCookie settingsAccessibility statement
        Huwise Terms & Conditions Cookie settings