REDCap has 5,900+ partner institutions, 2.1 million users, and zero MCP servers. I fixed that. Now my AI assistant exports records, audits data quality, and generates codebooks while I drink coffee. — Allan Mugarura, creator

Why This Exists

REDCap is the backbone of clinical and translational research worldwide. But every interaction — exporting records, checking completeness, generating codebooks, managing surveys — requires clicking through the web UI. This MCP server lets AI assistants do it programmatically.

What makes this different from a REST wrapper:

• Data quality audit — completeness per field, blank required fields, statistical outliers (z-score), duplicate detection, all in one call
• Auto-generated codebook — human-readable data dictionary with choices, branching logic, validation rules
• Enrollment dashboard — breakdown by arm, event, DAG, completion status, with timeline
• Descriptive statistics — mean/median/SD/IQR for numeric fields, frequency tables for categorical
• Cross-tabulation — contingency tables between any two fields
• Event comparison — compare a field across longitudinal timepoints
• Smart search — REDCap filter logic through natural language

Tools (38)

| Category | Tools | |----------|-------| | Project | get_project_info, create_project | | Data Dictionary | get_metadata, import_metadata | | Records | export_records, import_records, delete_records, next_record_id, search_records, record_count | | Instruments | list_instruments, get_instrument_event_mappings, get_field_names | | Events | list_events, import_events | | Arms | list_arms, import_arms | | Users | list_users, import_users, list_user_roles, list_user_role_assignments | | DAGs | list_dags, import_dags | | Reports | get_report | | Surveys | get_survey_link, get_survey_queue_link, get_survey_return_code, list_survey_participants | | Files | export_file, delete_file | | Repeating | list_repeating_instruments | | Audit | get_audit_log | | Analytics | data_quality_audit, enrollment_stats, generate_codebook, describe_field, compare_events, cross_tabulate |

Quick Start

1. Get Your REDCap API Token

In your REDCap project: API → Request API Token (or ask your REDCap admin).

You also need your institution's REDCap API URL, which looks like:

https://redcap.youruni.edu/api/

2. Install

npm install -g @masuratech/redcap-mcp-server

3. Register with Claude Code

claude mcp add redcap \
  --env REDCAP_API_URL=https://redcap.youruni.edu/api/ \
  --env REDCAP_API_TOKEN=YOUR_API_TOKEN_HERE \
  -- npx @masuratech/redcap-mcp-server

4. Use

Start a Claude Code session and ask things like:

• "Show me the data dictionary for my REDCap project"
• "Export all records where age > 18"
• "Run a data quality audit on the demographics form"
• "Generate a codebook for the project"
• "How many participants have completed the baseline survey?"
• "Cross-tabulate sex by treatment arm"
• "Compare depression scores across timepoints"
• "Get the audit log for record 101"
• "Generate a survey link for record 205"

Configuration

| Environment Variable | Required | Default | Description | |---------------------|----------|---------|-------------| | REDCAP_API_URL | Yes | — | Your REDCap instance API endpoint | | REDCAP_API_TOKEN | Yes | — | Project-level API token |

Alternative variable names REDCAP_URL and REDCAP_TOKEN are also accepted.

Multiple Projects

Register multiple MCP servers with different tokens:

claude mcp add redcap-study1 \
  --env REDCAP_API_URL=https://redcap.youruni.edu/api/ \
  --env REDCAP_API_TOKEN=TOKEN_FOR_STUDY_1 \
  -- npx @masuratech/redcap-mcp-server

claude mcp add redcap-study2 \
  --env REDCAP_API_URL=https://redcap.youruni.edu/api/ \
  --env REDCAP_API_TOKEN=TOKEN_FOR_STUDY_2 \
  -- npx @masuratech/redcap-mcp-server

Analytics Tools

Data Quality Audit

The data_quality_audit tool runs a comprehensive check in one call:

• Completeness: % filled per field, worst fields highlighted
• Required fields: identifies blank required-field entries
• Outliers: z-score based detection on numeric fields (configurable threshold)
• Duplicates: detects records with matching name/email/DOB/phone fields

Codebook Generation

The generate_codebook tool produces a structured, human-readable data dictionary:

• All forms with their labels
• Every field: name, label (HTML stripped), type, validation, choices, branching logic, notes
• Required field markers
• Choice code-label pairs for dropdowns/radios/checkboxes

Cross-Tabulation

The cross_tabulate tool generates contingency tables between any two categorical fields with row totals, column totals, and grand total.

Development

git clone https://github.com/masuratech/redcap-mcp-server.git
cd redcap-mcp-server
npm install
npm run dev    # watch mode
npm run build  # one-shot compile
npm start      # run the server

Security

• API tokens are passed via environment variables, never hardcoded
• All communication uses your institution's existing REDCap HTTPS endpoint
• No data is sent to third-party services — everything stays between you and your REDCap instance
• The server runs locally on your machine

License

MIT — Masura Tech Ltd

Built with mass-produced coffee in Kampala by Masura Tech.