MCP Tool Reference

The Quell connector exposes 8 tools. All tools read from synced reports — never from your local disk.

Tool 1 — list_projects

List all synced Quell projects for your account.

Input: none

Claude uses this when: you ask "what projects does Quell know about?" or refer to a project by an ambiguous name.

Tool 2 — get_prs_report

Get the latest PRS report for a project.

Input:

{ "project_alias": "payments-service" }

Output: full sync payload — PRS, edge case counts, written/scaffolded/flagged items.

Claude uses this when: you ask about PRS, test counts, or overall status.

Tool 3 — get_prs_history

Get PRS trend over time (last N runs).

Input:

{ "project_alias": "payments-service", "last_n_runs": 10 }

Claude uses this when: you ask "is my PRS improving?", "when did we drop below 80?".

Tool 4 — get_written_tests

Get WRITTEN test metadata, optionally filtered by confidence tier.

Input:

{ "project_alias": "payments-service", "tier": "MEDIUM" }

Tier values: HIGH (85 and above), MEDIUM (60–84), LOW (below 60), or omit for all.

Claude uses this when: you ask "which tests should I review?", "show me low-confidence tests".

Tool 5 — get_flagged_items

Get all FLAGGED items with reasons and file:line locations.

Input:

{ "project_alias": "payments-service" }

Claude uses this when: you ask "why can't these be auto-tested?", "what are my blind spots?".

Tool 6 — get_scaffolded_items

Get all SCAFFOLDED stubs with age in days (used for PRS penalty tracking).

Input:

{ "project_alias": "payments-service" }

Claude uses this when: you ask "which stubs have I not finished?", "am I losing PRS points?".

Tool 7 — get_badge_url

Get the SVG badge URL for a project's current PRS.

Input:

{ "project_alias": "payments-service" }

Output:

{
  "badge_url": "https://quell.buildsbyshashank.tech/badge/payments-service.svg",
  "markdown": "![Quell PRS](https://quell.buildsbyshashank.tech/badge/payments-service.svg)",
  "prs": 71,
  "tier": "Review Needed"
}

Claude uses this when: you ask "give me the badge for my README".

Tool 8 — reproduce_via_cloud

Trigger quell reproduce via the cloud API. Requires Pro or Team and the quell:reproduce:write OAuth scope.

Input:

{
  "project_alias": "payments-service",
  "bug_description": "payment of amount=-1 should raise ValueError but doesn't",
  "function_signature": "def process_payment(amount: float, currency: str) -> dict:",
  "docstring": "Raises: ValueError: if amount <= 0"
}

Only function_signature and docstring are sent to the LLM — never the function body. The generated test still runs all 5 gates server-side before being returned.

Claude uses this when: you paste a bug description and say "write a test for this".

Rate limits