CodeInspector Now Works Inside Claude: Scan and Fix Code in One Conversation

You can now run CodeInspector scans without leaving Claude. Connect it once as a connector, then ask in plain English: "scan this file", "scan our repo", "show me only the critical issues". Claude runs the scan, reads the results, and helps you fix what it found in the same conversation.

Video explainer (1:20) · Watch with transcript

Why we built it

CodeInspector started from years of reviewing code on a fintech team. The bugs that hurt in production were rarely the ones reviewers argued about. They were the quiet ones: a key committed by accident, an exception swallowed in a payments path.

Until now, catching them meant switching tabs: run a scan, read a report, come back to your editor or AI assistant, and explain the problem again. The Claude integration removes that loop.

What Claude can do with CodeInspector

The connector gives Claude six tools. You never call them by name; Claude picks the right one from what you ask.

  • scan_code: scans a single file or snippet and returns its issues right away (uses credits).
  • scan_repository: starts a background scan of a public repo, or a private repo you have connected in CodeInspector (uses credits).
  • get_scan_status: checks whether a repo scan has finished and summarises the issue counts.
  • get_scan_issues: lists a finished scan's issues, optionally filtered by severity.
  • list_recent_scans: shows your latest scans.
  • get_account: shows which account is connected and your remaining credits.

Set it up in about 30 seconds

In claude.ai, Claude Desktop or the Claude mobile app, add CodeInspector as a custom connector:

  1. Sign up at codeinspector.app (new accounts get 100 free credits)
  2. In Claude, open Settings → Connectors → Add custom connector
  3. Name it CodeInspector and paste https://mcp.codeinspector.app/mcp
  4. Click Add and sign in to CodeInspector
  5. Ask: "How many CodeInspector credits do I have left?"

Using Claude Code instead?

Add the server from your terminal with your CodeInspector API key, then use it from any project.

claude mcp add --transport http codeinspector https://mcp.codeinspector.app/mcp \
  --header "Authorization: Bearer $CODEINSPECTOR_API_KEY"

A real example: scanning a payments file

Here is an excerpt from a small, deliberately sloppy payments.py, the kind of code that slips through on a busy day. We pasted it into Claude and asked: "Can you scan payments.py before I open a pull request?"

API_SECRET = "sk_live_51HxQ9fintechDemoKey"

def transfer(from_acct, to_acct, amount):
    try:
        balance = get_user_balance(from_acct)[0]
        if balance > amount:
            print("Transferring", amount, "using key", API_SECRET)
            return True
    except:
        pass
    return False

What CodeInspector found

Claude called the scan_code tool. A few seconds later it came back with three issues, using 5 credits:

  • Critical — Hardcoded secret (line 4): load secrets from environment variables or a secret store.
  • High — Bare except clause (line 22): catch specific exception types. A bare except also catches SystemExit and KeyboardInterrupt, so every real failure in a payments path disappears silently.
  • Low — Missing logging (line 1): log exceptions with context.

From findings to a fix

Because the results land inside Claude, the next step is just "fix these". Claude rewrote the function to read the secret from an environment variable, catch only the errors it expects, and log failures. The fix also stops printing the API key to stdout, which the original did on every successful transfer.

import os, logging, sqlite3

log = logging.getLogger(__name__)
API_SECRET = os.environ["PAYMENTS_API_KEY"]

def transfer(from_acct, to_acct, amount):
    try:
        balance = get_user_balance(from_acct)[0]
        if balance > amount:
            log.info("Transfer %s -> %s", from_acct, to_acct)
            return True
    except (sqlite3.Error, TypeError) as e:
        log.exception("Transfer failed: %s", e)
    return False

Scanning a whole repository

For a full codebase, give Claude the repo: "Scan github.com/your-org/payments-api and show me only the critical issues." Repo scans run in the background, so Claude handles the steps for you.

Private repos work too, as long as you have connected them in your CodeInspector account first. Every scan also gets a full report on codeinspector.app that you can share with your team.

  1. scan_repository starts the scan
  2. get_scan_status checks until it is done
  3. get_scan_issues pulls the critical findings with file, line and fix

Try it today

CodeInspector scans C#, Java, Python, C++, JavaScript and TypeScript. Every new account gets 100 free credits, enough to try the Claude integration on real code. Paid plans start at $19 a month for individuals, $49 for teams and $199 for enterprise, and all paid plans are 20% off until the end of November 2026.

Watch the 80-second demo above, then connect CodeInspector to Claude and scan your next pull request before anyone else reviews it.

Related