The Code Registry – New release v1.3.5 – 12th February 2026

This has been in the works for a while — our official MCP server. Using natural language, anyone can create an account, analyze code up to our free tier, and see the results. You can ask your AI agent to present the results in a nice way with charts and graphs, or however else you want to use the data — making it easy to incorporate our enterprise-grade code intelligence into your AI-based development flow.

For example, you can tell your normal coding agent to “Use The Code Registry to summarise security and tech debt after each task is complete.”

This is our official public MCP server, which explains how it works: github.com/The-Code-Registry/mcp-server

And for your AI systems, you simply need to add an MCP server (sometimes called a “Connector”) with this URL: https://integrator.app.thecoderegistry.com/api/ai/router

With this release we’re opening up our dedicated code remediation service to the world. So now not only do we help businesses understand what’s going on with their code, we can instantly quote for remediating any issues found, and then work with the customer on resolving them — all with our new dedicated code remediation team of developers.

Our remediation team has experienced engineers in all sorts of technologies, libraries, and languages. Their sole job is to work with our customers to improve the quality, security, and technical debt of their digital assets. Our mission is to help all businesses globally better understand their code — and now we can help those businesses improve it!

We’ve been talking to customers using our autonomous code exploration agent Code IQ, and some of the real-world use cases are fascinating — from non-technical CEOs researching bugs and translating them for their dev team, to R&D tax specialists finding evidence of proprietary work within a codebase for tax claims.

As part of this feedback, we’ve discovered some improvements and bug fixes. The main one: we’ve increased our rate limits with the underlying LLM model by 12–13x, meaning Code IQ can explore even larger codebases simultaneously than before.

We’ve made big improvements to the stability of our security scanning feature — our most computer-intensive function, and one we’ve spent a lot of time tweaking for performance and resource usage since launch.

We’ve now implemented an architecture design pattern that can handle extremely large codebases: a “bigger brother” security scanner with beefed-up resources and an almost unlimited time limit. When the normal scanner repeatedly has issues on a large codebase, it hands off the scan to the larger function. This costs more compute but is essential for handling codebases of any size — the customer sees no errors, and the scan keeps going until it’s complete, no matter the size.

• Fixed a bug in very large files hosted on Google Drive. Google Drive was previously showing a warning page when files were over 100MB; we now detect this and work around it properly.
• Better support for Dropbox file archive URLs with mixed parameter orders.
• Fixed an issue generating project-level combined reports when all code vaults are file-based.
• Fixed a bug in the local code agent due to specific security scanner versions.
• Fixed a bug in the local code agent where it would silently fail if http/https was in the web endpoint used.
• Fixed a bug in security scanner where invalid XML in the scan results (rare) was not caught or retried.
• Fixed a bug in the security page where CRITICAL issues wouldn’t export into the CSV or JSON files.