Claude Error Collector FAQs

Question 1

How does it prevent Claude from making the same coding errors twice?

Accepted Answer

When you correct Claude during a session, the Error Collector captures that lesson as a rule. Claude then reads these rules from your `CLAUDE.md` file in future interactions, ensuring it applies the learned corrections and avoids repeating those specific errors.

Question 2

How does the tool handle duplicate rules?

Accepted Answer

It uses both server-side and semantic duplicate detection. Before writing, the server checks for case-insensitive matches. Additionally, Claude is instructed to review for semantic equivalence; if a new rule is similar to an existing one, it may ask you whether to add, consolidate, or skip.

Question 3

Does it support both global and project-specific rules?

Accepted Answer

Absolutely. The Claude Error Collector intelligently resolves `CLAUDE.md` files. It will write and read rules from a project-specific `CLAUDE.md` if detected in your current working directory, or fall back to a global `~/.claude/CLAUDE.md` for general rules.

Question 4

What is the Claude Error Collector?

Accepted Answer

The Claude Error Collector is an MCP server for Claude Code that automates learning from your coding corrections. It records your feedback as specific rules in `CLAUDE.md` to prevent Claude from repeating past mistakes.

Question 5

Can I manage the rules learned by Claude?

Accepted Answer

Yes, the tool provides a suite of features to manage your learned rules. You can list all rules, update existing ones, and delete rules by index or matching text. It also includes intelligent duplicate detection.

Claude Error Collector

Claude Error Collector

主な機能

ユースケース

主な機能

ユースケース