ROBINS-I V2 MCP — Claude Code MCP server: what it does, how to install
Evaluates risk of bias in non-randomized cohort studies using ROBINS-I V2 with deterministic quote resolution and scoring graphs.
claude mcp add robins-i-mcp -- uvx robins-i-mcpWhat ROBINS-I V2 MCP does
The server runs structured risk of bias assessments for observational cohorts and target trial emulations. Instead of letting an LLM generate arbitrary scores, it limits the model to answering signalling questions with verbatim excerpts. A deterministic engine verifies each excerpt against character positions in the document bundle, while six domain algorithms compute the verdicts through explicit graphs.
Evaluating confounding in domain 1 is hard-blocked until you provide external confounders via set_prespecified_confounders. You must also define protocol adherence via the C4 parameter in specify_result, which chooses between intention-to-treat and per-protocol pathways. Omission of supplementary files causes missing-information judgements, because authors rarely put target trial emulation details in main article bodies.
What it helps with
- Calculate risk of bias metrics via deterministic ROBINS-I V2 graphs
- Validate extracted paper excerpts against exact character offsets
- Verify absent methodological details through automated manuscript searches
- Export multi-study assessment records into CSV tables for
robvis
Example prompts for the agent
- Ingest the paper using
parse_pmcidwith PMC1234567 and inspect its structure - Supply our baseline confounder list using
set_prespecified_confoundersand run domain 1 - Compile all generated session records into a
robvisCSV viaexport_robvis
How to install ROBINS-I V2 MCP
Platform CLI
claude mcp add robins-i-mcp -- uvx robins-i-mcpuvx runs the command named after the package; if the README names another one, use uvx --from robins-i-mcp <command>
Platform CLI
codex mcp add robins-i-v2-mcp -- uvx robins-i-mcpuvx runs the command named after the package; if the README names another one, use uvx --from robins-i-mcp <command>
Config file
{
"mcpServers": {
"robins-i-v2-mcp": {
"command": "uvx",
"args": [
"robins-i-mcp"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from robins-i-mcp <command>
Config file
{
"servers": {
"robins-i-v2-mcp": {
"type": "stdio",
"command": "uvx",
"args": [
"robins-i-mcp"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from robins-i-mcp <command>
Platform CLI
gemini mcp add robins-i-v2-mcp uvx robins-i-mcpuvx runs the command named after the package; if the README names another one, use uvx --from robins-i-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 ROBINS-I V2 MCP
How do I install ROBINS-I V2 MCP in Claude Code?
Run the command claude mcp add robins-i-v2-mcp -- uvx robins-i-mcp. The assessment tools will immediately register in your agent session.
What does ROBINS-I V2 MCP do?
It evaluates risk of bias in cohort designs using the ROBINS-I V2 framework. The server verifies model-supplied text evidence against original documents and computes domain verdicts via explicit logic trees.
Why does domain 1 require a separate setup step?
The engine pauses domain 1 until you run set_prespecified_confounders. ROBINS-I V2 requires benchmarking study adjustments against the reviewer's protocol rather than the authors' own covariate tables.
