Guides
Real examples of how developers use MCP tools in daily workflows. Each scenario shows which tools are called, in what order, and what the result looks like.
SCENARIO
You just joined a team and need to understand a 200-file project quickly.
TOOLS USED
get_project_context → explain_architecture → find_code
HOW IT WORKS
First, get_project_context gives you the full stack snapshot: framework, dependencies, scripts, env vars. Then explain_architecture generates a diagram showing how services connect. Finally, find_code lets you search for specific patterns like auth middleware or database queries.
RESULT
You understand the project in 10 minutes instead of spending a day reading files manually.
SCENARIO
Dependabot opened a PR to bump Next.js from 15 to 16. You need to know what breaks.
TOOLS USED
check_package (task=upgrade) → check_test
HOW IT WORKS
check_package with task=upgrade returns every breaking change between versions, CVEs patched, deprecated APIs, and migration steps. Then check_test tells you exactly which files in your project are affected by those breaking changes.
RESULT
You know the full blast radius before merging, with a file-by-file action plan.
SCENARIO
Users report 500 errors on the /api/checkout endpoint. You need to trace the cause.
TOOLS USED
keep_thinking → find_code → read_code → audit_headers
HOW IT WORKS
Start with keep_thinking to structure your investigation and track hypotheses. Use find_code to locate the checkout handler. Use read_code to examine the implementation. If it might be a header/CORS issue, audit_headers checks the deployed endpoint.
RESULT
Systematic debugging that tracks what you checked and what you ruled out, preventing circular investigation.
SCENARIO
A tutorial recommends installing a package you have never heard of.
TOOLS USED
check_package
HOW IT WORKS
check_package queries the live registry and reports: version status, weekly downloads, maintainer count, postinstall hooks, typosquat risk, and whether the package name matches known hallucinated/fake names from AI tools.
RESULT
You catch supply-chain attacks and typosquats before they enter your project.
SCENARIO
You want to extract the auth logic into its own module but are not sure what will break.
TOOLS USED
check_test → explain_architecture → find_code
HOW IT WORKS
check_test identifies every file that imports or calls the auth functions, with severity ratings (breaking, likely_affected, possibly_affected). explain_architecture shows the auth flow visually. find_code confirms all call sites.
RESULT
You have a complete impact map before writing a single line of code.
SCENARIO
A teammate submitted a PR that adds a new API endpoint. You need to review it thoroughly.
TOOLS USED
read_code → check_package → audit_headers
HOW IT WORKS
read_code extracts the new endpoint implementation. If new dependencies were added, check_package verifies they are safe. After the PR is deployed to staging, audit_headers confirms security headers are correct.
RESULT
Thorough review covering code quality, dependency safety, and deployment security.
Ready to try these workflows? Follow the Quickstart guide to connect your editor in under 5 minutes.