Target MCP server
{
"mcpServers": {
"blackswancausallabs-target": {
"command": "uvx",
"args": [
"target-mcp"
],
"env": {
"ANTHROPIC_API_KEY": "sk-ant-..."
}
}
}
}{
"mcpServers": {
"blackswancausallabs-target": {
"command": "uvx",
"args": [
"target-mcp"
],
"env": {
"ANTHROPIC_API_KEY": "sk-ant-..."
}
}
}
}{
"mcpServers": {
"blackswancausallabs-target": {
"command": "uvx",
"args": [
"target-mcp"
],
"env": {
"ANTHROPIC_API_KEY": "sk-ant-..."
}
}
}
}{
"servers": {
"blackswancausallabs-target": {
"command": "uvx",
"args": [
"target-mcp"
],
"env": {
"ANTHROPIC_API_KEY": "${input:anthropic_api_key}"
}
}
},
"inputs": [
{
"id": "anthropic_api_key",
"type": "promptString",
"description": "Anthropic api key",
"password": true
}
]
}{
"mcpServers": {
"blackswancausallabs-target": {
"command": "uvx",
"args": [
"target-mcp"
],
"env": {
"ANTHROPIC_API_KEY": "sk-ant-..."
}
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsoninstall uvx target-mcpversion 0.2.0
What it does
An MCP server that scores how completely a target trial emulation study reports what the TARGET reporting guideline requires. It processes manuscripts (PDF, text, docx, or fetched from Europe PMC) and produces structured item-by-item assessments with evidence linked to specific document locations. Built for researchers, reviewers, and corpus-scale studies evaluating observational studies that emulate randomized trials.
- Parse and assess a manuscript or bundle of files against the 39 TARGET checklist items, with verdicts validated against required evidence spans.
- Fetch open-access papers by PMCID from Europe PMC and automatically retrieve their supplementary files, then score completeness of TARGET reporting.
- Aggregate assessments across multiple studies to produce per-item completeness rates and run validation studies with blind coding and inter-rater agreement metrics (Cohen's κ, Gwet's AC1).
Tools this server exposes
| Tool | Does |
|---|---|
| parse_manuscript | Parse a manuscript file (PDF, text, or docx) into a structured section map with source-tagged spans and optional supplement files. |
| assess_manuscript | Score all applicable TARGET checklist items against a parsed manuscript, returning prompts and schema for scaffold mode or making pinned model calls for judge m |
| submit_scaffold_verdicts | Validate and record verdicts submitted for a manuscript assessment, ensuring required evidence is resolved to document spans. |
| render_checklist | Project a finalized assessment onto the published TARGET checklist form with item wording, locations, verdicts, and evidence. |
| render_checklist_docx | Write a finalized assessment as a submission-ready Word file matching the TARGET checklist format. |
| parse_pmcid | Fetch an open-access article from Europe PMC by PMCID, retrieve JATS main text and supplementary files, and merge them into a single bundle. |
| get_checklist | Introspect and return the TARGET specification, including all 39 scoreable leaf subitems with intent, verdict boundaries, and applicability rules. |
| aggregate_corpus | Roll multiple manuscript assessments into per-item completeness rates with coverage denominators. |
| build_coding_sheet | Generate a blind human coding sheet for validation studies. |
| validate_against_gold | Compare assessments against a gold-standard coding sheet and compute per-leaf agreement metrics including raw agreement, Cohen's κ, Gwet's AC1, sensitivity, and |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.