ARQERA MCP Server
Let AI assistants interact with ARQERA natively. Governance, agents, evidence, and compliance — all accessible via the Model Context Protocol.
What is MCP?
The Model Context Protocol — a standard for AI assistant integrations
The Model Context Protocol (MCP) is an open standard that lets AI assistants — Claude, ChatGPT, Copilot, and others — interact with external tools and services directly. Instead of copy-pasting data between your AI assistant and ARQERA, the MCP server gives your assistant native access to governance, compliance, evidence, and agent management.
Your AI assistant becomes a governance-aware operator. It can evaluate actions before executing them, check compliance status, query the audit trail, and manage approvals — all through natural conversation.
Governance Evaluation
AI assistants can evaluate actions against your governance policies before executing them, ensuring compliance in real time.
Agent Execution
Execute, approve, and reject AI actions directly from your AI assistant. Full human-in-the-loop support.
Evidence & Audit
Query the evidence chain, export audit reports, and verify integrity — all from a natural language conversation.
Compliance Checks
Run SOC 2, GDPR, and EU AI Act compliance checks and get instant status reports inside your AI chat.
Installation
Install the ARQERA MCP server package
npm
$ npm install @arqera/mcp-servernpx (no install)
$ npx @arqera/mcp-serverConfiguration for Claude Desktop
Add ARQERA to your Claude Desktop configuration
claude_desktop_config.json
{
"mcpServers": {
"arqera": {
"command": "npx",
"args": ["@arqera/mcp-server"],
"env": {
"ARQERA_API_KEY": "ak_live_your_api_key_here",
"ARQERA_BASE_URL": "https://app.arqera.io"
}
}
}
}Add this to your Claude Desktop configuration file. On macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Configuration for Claude Code
Add ARQERA to your Claude Code project
.mcp.json (project root)
{
"mcpServers": {
"arqera": {
"command": "npx",
"args": ["@arqera/mcp-server"],
"env": {
"ARQERA_API_KEY": "ak_live_your_api_key_here"
}
}
}
}Place this file in your project root. Claude Code will automatically detect and connect to the ARQERA MCP server when you open the project.
Available Tools
All MCP tools exposed to AI assistants
| Tool | Description |
|---|---|
| governance.evaluate | Evaluate an action against governance policies. Returns tier, risk score, and policy matches. |
| ara.execute | Execute an AI action through the governance engine. AUTO/SOFT actions run immediately. |
| ara.approve | Approve a pending HARD-tier action by ID. |
| ara.reject | Reject a pending action with an optional reason. |
| evidence.list | List evidence artifacts. Filter by date range, type, or actor. |
| evidence.export | Export the evidence chain as JSON, CSV, or PDF. |
| compliance.check | Run a compliance check against a specific framework. |
| integrations.list | List all connected integrations with status and health. |
| agents.list | List all deployed agents with status and capabilities. |
| diagnostics.health | Get a system health summary including service status and performance metrics. |
Example Conversations
How AI assistants use ARQERA tools in practice
Check if our SOC 2 compliance is up to date.
I'll run a SOC 2 compliance check for you.
compliance.check({ framework: "soc2" })Your SOC 2 compliance is at 94% coverage. 47 of 50 controls are passing. 3 controls need attention: CC6.1 (logical access), CC7.2 (change management), CC8.1 (incident response).
Send an email to the team about the quarterly review.
Let me evaluate this action through governance first.
governance.evaluate({ action: "email.send", context: { recipients: "[email protected]", subject: "Quarterly Review" } })This action is classified as HARD tier (external communication). It requires your approval before sending. Shall I create the pending approval?
Show me what happened in the audit trail today.
I'll pull today's evidence artifacts.
evidence.list({ since: "2026-02-19", limit: 20 })Today there are 14 evidence artifacts: 6 governance evaluations, 4 agent executions, 2 compliance checks, and 2 integration events. The most recent was a data export approved by you at 14:32.
Authentication
Secure your MCP server with an API key
Getting an API key
- Log in to app.arqera.io
- Go to Settings → API Keys
- Click Create API Key
- Select scopes:
governance:read,governance:write,evidence:read,compliance:read - Copy the key and set it as the
ARQERA_API_KEYenvironment variable
Environment variable
# Set in your shell profile or CI/CD environment export ARQERA_API_KEY="ak_live_your_api_key_here"
Security
How ARQERA keeps AI-initiated actions safe
Tool Scoping
Each API key has specific scopes. An AI assistant can only access the tools permitted by its key. Read-only keys cannot execute actions.
Tenant Isolation
All MCP operations are tenant-scoped. An AI assistant connected to Tenant A cannot access Tenant B's data, policies, or evidence.
Audit Trail
Every action initiated through MCP is recorded in the evidence chain with the actor marked as the AI assistant and the authorizing user.
Human-in-the-Loop
HARD-tier actions always require human approval, even when initiated by an AI assistant. The governance engine cannot be bypassed via MCP.
Give your AI assistant superpowers
Connect ARQERA to Claude, ChatGPT, or any MCP-compatible AI assistant in under 5 minutes.