TARGET Checklist MCP — Claude Code MCP server: TTE report scoring
Scores how fully a target trial emulation study reports each item the TARGET guideline asks for, and returns a per-subitem matrix with quotes.
claude mcp add target-mcp -- uvx target-mcpWhat TARGET Checklist MCP does
The server ingests a manuscript in PDF, text or docx, optionally bundled with supplements, and maps its sections with source tags. It then scores 39 leaf subitems carved out of the 21 published TARGET items. The spec lives in a versioned YAML with verdict boundaries and applicability rules, and it is validated on load.
Default scaffold mode keeps the agent in the loop as the scorer: the server hands it a prompt and schema, then checks the submitted verdicts for leaf coverage, allowed vocabulary and verbatim evidence tied to resolved spans. Judge mode runs headless with its own pinned model call, so a batch pass over a corpus stays reproducible. Output goes to the published checklist form or to a Word file.
Read the tally as reporting completeness only. There is no pass/fail verdict, and a matrix of verdicts is not a claim of TARGET compliance. Install with uvx; no repo clone needed. The optional fan-out skill speeds up scoring by running leaves in parallel, but it needs a client that can spawn subagents — Codex and other non-Claude clients skip it and fall back to the single-pass path.
What it helps with
- score a manuscript against every TARGET item
- produce a checklist with quoted evidence per row
- measure item-level completeness across a batch of papers
- generate a blind coding sheet for manual review
- compare leaf verdicts to a gold standard and compute agreement
Example prompts for the agent
- Parse
study.pdfwithsupplement.pdfand score it against TARGET - Fetch an open-access paper by PMCID and assess reporting completeness
- Render the finished checklist as a docx file
How to install TARGET Checklist MCP
Platform CLI
claude mcp add target-mcp -- uvx target-mcpuvx runs the command named after the package; if the README names another one, use uvx --from target-mcp <command>
Platform CLI
codex mcp add target-checklist-mcp -- uvx target-mcpuvx runs the command named after the package; if the README names another one, use uvx --from target-mcp <command>
Config file
{
"mcpServers": {
"target-checklist-mcp": {
"command": "uvx",
"args": [
"target-mcp"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from target-mcp <command>
Config file
{
"servers": {
"target-checklist-mcp": {
"type": "stdio",
"command": "uvx",
"args": [
"target-mcp"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from target-mcp <command>
Platform CLI
gemini mcp add target-checklist-mcp uvx target-mcpuvx runs the command named after the package; if the README names another one, use uvx --from target-mcp <command>
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about TARGET Checklist MCP
How do I install TARGET Checklist MCP in Claude Code?
Run claude mcp add target-mcp -- uvx target-mcp, then restart or reconnect the client and check that 11 tools show up.
What does TARGET Checklist MCP do?
It parses a manuscript, scores it across 39 leaf subitems of the TARGET guideline and builds a matrix with verbatim quotes tied to sections. The result can be exported as the checklist form or a Word file.
Is this a study quality check?
No. Only reporting completeness is measured: how fully the text covers what the checklist requires. The tool issues no judgement on study quality.
