diagnose_repository
Analyze intake signals and return architecture, health, and intent before anything changes.
Example: “Run diagnose_repository on this repo — stabilize architecture before onboarding.”
A R K I T E C T M C P
Desktop-first architecture reasoning with MCP-native tools for Cursor. Diagnosis, pattern intelligence, and guidance grounded in your actual repo.
Arkitect is free for the first 1,000 people who claim a spot. After that, membership pricing takes over — claim yours now to lock in free access.
Loading live claim count…
No credit card required. One free spot per visitor.
Click a card to unlock its tools. Click again to lock.
Diagnosis is the entry point — scan the repo, then decide. Nothing refactors until you ask.
diagnose_repository
Analyze intake signals and return architecture, health, and intent before anything changes.
Example: “Run diagnose_repository on this repo — stabilize architecture before onboarding.”
get_last_diagnosis
Reuse the latest diagnosis payload without scanning again.
Example: “Show me the last diagnosis result.”
list_diagnosis_strategies
See how Arkitect continues healthy structure, reports drift, or defers a refactor.
Example: “What diagnosis strategies does Arkitect apply?”
suggest_requirement_tags
Suggest scope tags from the repo so diagnosis stays pointed at the right work.
Example: “Suggest requirement tags for this repo.”
Pattern intelligence — the catalog behind the algorithm. Same vocabulary as the architecture guide.
get_pattern_intelligence
Deep GoF data: intent, applicability, steps, pros/cons, and relations.
Example: “Get pattern intelligence for observer.”
recommend_patterns
Recommend patterns for this repo, with deferred options and an ADR summary.
Example: “Recommend patterns with complexityProfile balanced.”
list_design_principles
SOLID plus general OO principles Arkitect uses when scoring structure.
Example: “List the design principles Arkitect uses.”
list_design_patterns
The full catalog grouped by creational, structural, and behavioral families.
Example: “List design patterns available in the Arkitect catalog.”
Run verify and tests from the repo root — structured pass/fail the agent can act on.
verify_codebase
Full pipeline: lint, build, typecheck, test.
Example: “Verify the codebase at this path.”
run_tests
Unit and integration tests only — lighter than a full verify.
Example: “Run tests in this repo.”
run_test_suite
Scope a suite: unit, integration, or all.
Example: “Run the integration test suite.”
Report-only. Arkitect ranks techniques — it does not rewrite until you explicitly ask.
analyze_refactoring_opportunities
Rank Refactoring Guru techniques against structural smells. Reports only.
Example: “Analyze refactoring opportunities with explicit refactor intent.”
list_refactoring_techniques
The technique catalog grouped by category, with reference URLs.
Example: “List refactoring techniques for simplifying-conditionals.”
BEHIND THE ALGORITHM
Diagnosis maps your repo onto design patterns, architecture styles, and SOLID — then reports before it refactors. Open the full guide on its own page.
OPEN THE GUIDEA R K I T E C T M C P
Arkitect scans your repo and surfaces the real architecture, health, and intent before anything changes.
Detected structure keeps flowing automatically once confidence is high — no forced rewrites.
Drifting or spaghetti structure is reported clearly. Refactors only happen with explicit intent.
{
"mcpServers": {
"arkitect-mcp": {
"command": "node",
"args": [
"packages/mcp-server/dist/stdio.js"
]
}
}
}
Drop this into your Cursor MCP settings and restart the client.
Lock in free access while spots remain
Point Cursor at the Arkitect stdio server
Drop the config into your MCP settings
Ask your agent to scan repo structure
Claim your spot while free access lasts. Membership pricing arrives after the free cohort — early claimants keep free access forever.
CLAIM YOUR SPOTYou always connect your own provider API keys in the desktop app — or use Cursor's built-in AI when Arkitect runs as MCP in chat. Provider choice does not change how Arkitect reasons about your repo.
Fast, cost-efficient. Default recommendation for Arkitect reasoning tasks.
Claude models for deep architectural analysis and complex reasoning.
GPT models for pattern recognition and code generation tasks.
Google's models for multi-modal analysis and large context windows.
Ultra-fast inference for real-time diagnosis and quick scans.
Bring your own endpoint — Ollama, LM Studio, or any OpenAI-compatible API.
"Arkitect caught architectural drift in our monorepo that three code reviews missed. The diagnosis-first approach saved us weeks of refactoring in the wrong direction."
Marcus Kim
STAFF ENGINEER
"Finally, an MCP tool that actually understands software architecture instead of just pattern-matching code. The vertical slice detection alone is worth it."
Ana Santos
TECH LEAD
"Set it up in Cursor in under five minutes. The pattern recommendations are spot-on — it's like having a senior architect reviewing every decision."
James Doyle
INDIE DEVELOPER
FREE TIER
$0
first 1,000 users
PRO
—
coming after free cohort
TEAM
—
coming after free cohort
Follow along or reach out if you need help — bugs, feature requests, or feedback. Every message gets read.