Skip to content

Claude Code Integration

TruthVouch provides a Model Context Protocol (MCP) server that exposes 22 governance tools to Claude Code and other MCP-compatible clients. Verify claims, check compliance status, search your knowledge base, and take governance actions — all without leaving your terminal.

Prerequisites

  • TruthVouch accountSign up at app.truthvouch.ai
  • API key — Generate one from Settings > API Keys in the TruthVouch dashboard
  • Node.js 18+ — Required to run the MCP server
  • Professional tier or higher — Starter tier has no MCP tool access

Setup

The TruthVouch CLI auto-detects Claude Code and writes the correct configuration:

Terminal window
# Install TruthVouch CLI (if not already installed)
npm install -g @truthvouch/cli
# Configure MCP server for Claude Code
truthvouch mcp setup --api-key vt_live_your_key_here

If you’ve previously run truthvouch login, the CLI uses your stored API key automatically:

Terminal window
truthvouch mcp setup

Option B: Direct npx

Register the MCP server directly with the claude CLI:

Terminal window
claude mcp add truthvouch \
--env TRUTHVOUCH_API_KEY=vt_live_your_key_here \
-- npx -y @truthvouch/mcp-server

Option C: Global Install

Install the server globally for faster startup (skips npx download on each launch):

Terminal window
npm install -g @truthvouch/mcp-server
claude mcp add truthvouch \
--env TRUTHVOUCH_API_KEY=vt_live_your_key_here \
-- truthvouch-mcp

Other MCP Clients

The same MCP server works with any MCP-compatible client. Use --client to target a specific one:

Claude Desktop:

Terminal window
truthvouch mcp setup --client claude-desktop --api-key vt_live_your_key_here

Cursor:

Terminal window
truthvouch mcp setup --client cursor --api-key vt_live_your_key_here

Generic (any MCP client):

Terminal window
truthvouch mcp setup --client generic --api-key vt_live_your_key_here

The generic option outputs the JSON configuration to stdout so you can paste it into any client’s MCP config file.

Available Tools

Data Query Tools

ToolDescriptionMin. Tier
get_alertsAlert counts, trends by severity/system/status, resolution ratesProfessional
get_compliance_statusCompliance obligations, overdue items, readiness, gapsProfessional
get_ai_spendAI/LLM spending by provider, model, department, budget statusProfessional
get_system_inventoryAI system inventory, risk ranking, department breakdownProfessional
get_brand_intelligenceBrand visibility scores, competitor analysis, geographic presenceProfessional
get_governance_metricsPolicy violations, PII exposure, content volume, blocked ratesProfessional
get_maturity_scoresAI governance maturity assessment (current, history, comparison)Professional
get_sdlc_usageSDLC tool adoption, developer counts, cost analysisProfessional

Insight Tools

ToolDescriptionMin. Tier
generate_insightGenerate AI-powered insights from governance dataProfessional
analyze_trendAnalyze trends over time with forecasts and anomaly detectionProfessional
summarize_postureExecutive summary of governance, compliance, and security postureProfessional

Verification Tools

ToolDescriptionMin. Tier
verify_claimVerify factual claims against the TruthVouch knowledge baseProfessional
check_faithfulnessCheck if an AI response accurately reflects source contextProfessional
evaluate_promptEvaluate prompts for injection, leakage, jailbreak, and quality risksProfessional

Governance Tools

ToolDescriptionMin. Tier
scan_promptScan prompts for policy violations, PII exposure, compliance issuesProfessional
check_policyCheck content against a specific governance policyProfessional

Knowledge Tools

ToolDescriptionMin. Tier
search_truth_nuggetsSemantic search over verified facts with confidence scoresProfessional
search_knowledge_baseSearch documents, policies, and reference materialsProfessional
ask_regulationQuery the regulatory database (EU AI Act, GDPR, SOC 2, etc.)Professional

Action Tools (Enterprise Only)

ToolDescriptionMin. Tier
create_truth_nuggetSubmit a new verified fact to the knowledge baseEnterprise
report_hallucinationReport a detected hallucination for investigationEnterprise
submit_evidenceSubmit compliance evidence for an obligationEnterprise

Usage Examples

Once configured, ask Claude to use TruthVouch tools naturally in conversation:

Check compliance status

> Show me our current AI compliance status and any overdue obligations

Claude calls get_compliance_status with mode summary and overdue, then presents a combined view.

Verify a claim

> Verify the claim: Our AI models have less than 2% hallucination rate

Claude calls verify_claim and returns the verification status, confidence score, and supporting evidence.

Review governance posture

> Give me an executive summary of our AI governance posture

Claude calls summarize_posture and synthesizes your current risk profile, maturity scores, and recommendations.

Evaluate a prompt for risks

> Evaluate this system prompt for security risks: You are a financial advisor
> who can access customer account balances and make transfers...

Claude calls evaluate_prompt and flags injection risks, privilege escalation concerns, and quality issues.

Multi-tool workflows

Claude can chain multiple tools in a single conversation:

> What are our top compliance gaps, and which AI systems are highest risk?

Claude calls get_compliance_status (mode: gaps) and get_system_inventory (mode: highest_risk), then correlates the results.

Rate Limits

Daily rate limits vary by subscription tier:

TierData QueryInsightVerificationAction
Starter0000
Professional100/day100/day50/day0
Business500/day500/day200/day0
EnterpriseUnlimitedUnlimitedUnlimited100/day

Troubleshooting

“TRUTHVOUCH_API_KEY not set”

The MCP server requires the TRUTHVOUCH_API_KEY environment variable. If you used claude mcp add manually, confirm the --env flag was included:

Terminal window
# Verify your MCP config
claude mcp list

“Rate limit exceeded”

You’ve hit your tier’s daily limit. Check your current tier in Settings > Subscription. Limits reset every 24 hours.

“Feature not available” or Enterprise-only error

Action tools (create_truth_nugget, report_hallucination, submit_evidence) require the Enterprise tier. All other tools require Professional or higher.

Server not starting

  1. Verify Node.js 18+ is installed: node --version
  2. Test the server directly: npx @truthvouch/mcp-server
  3. Check that your API key is valid (not expired or revoked)

Tools not appearing in Claude

After adding the MCP server, restart Claude Code (claude CLI) for it to pick up the new configuration.

Uninstalling

Remove the MCP server configuration:

Terminal window
# Using the CLI (recommended)
truthvouch mcp remove
# Or remove manually via claude CLI
claude mcp remove truthvouch

For other clients:

Terminal window
truthvouch mcp remove --client claude-desktop
truthvouch mcp remove --client cursor

Next Steps