MCP Getting Started

The Model Context Protocol (MCP) lets AI assistants query your API traffic directly. Ask questions in natural language and get instant insights about connections, requests, errors, and data exposure.

What you'll accomplish:

• Get your authentication token

• Connect an MCP client to Qplane

• Run your first traffic query

Time to complete: 10 minutes

Prerequisites:

• An active Qplane account with at least one Qtap agent deployed

• An MCP client (Claude Code, Codex CLI, Cursor, or similar)

---

What MCP Enables

With MCP, you can ask your AI assistant questions like:

• "Show me all connections to Stripe in the last hour"

• "What errors occurred on the payments endpoint yesterday?"

• "Which endpoints are receiving the most traffic?"

• "Is there any PII detected in outbound traffic?"

The AI assistant translates your questions into structured queries against your Qplane data, giving you instant visibility without navigating dashboards or writing queries manually.

MCP provides read-only access to your traffic data. It cannot modify configurations or affect traffic flow.

---

Step 1: Get Your Token

1. Log in to app.qpoint.io

2. Navigate to Settings → Deployments

3. Click + New Registration Token

4. Name it something descriptive (e.g., "MCP Access")

5. Copy the generated Auth Token

Store your token securely. It provides access to your organization's traffic data.

---

Step 2: Connect Your MCP Client

Claude Code

claude mcp add qpoint --transport http https://pulse.qpoint.io/mcp/v1 \
  --header "Authorization: Bearer YOUR_TOKEN"

Verify the connection:

claude mcp list

Codex CLI

Add to ~/.codex/config.toml:

[mcp_servers.qpoint]
url = "https://pulse.qpoint.io/mcp/v1"
bearer_token_env_var = "QPOINT_TOKEN"

Set your token:

export QPOINT_TOKEN=your-token-here

Cursor

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "qpoint": {
      "transport": "http",
      "url": "https://pulse.qpoint.io/mcp/v1",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

Other Clients

Configure with:

• URL: https://pulse.qpoint.io/mcp/v1

• Transport: Streamable HTTP

• Header: Authorization: Bearer YOUR_TOKEN

---

Step 3: Your First Query

Once connected, try these starter queries:

Check Connection Status

"What endpoints have I connected to in the last hour?"

This verifies the MCP connection is working and shows recent traffic.

View Traffic Summary

"Give me a traffic summary for today"

Returns connection counts, request volumes, and error rates.

Find Errors

"Show me any 500 errors from today"

Lists requests that returned server errors.

---

Common Use Cases

Security & Compliance

"What external APIs are my applications calling that aren't from known vendors?"

"Show me requests that failed authentication (401/403 errors) in the past week"

"Is there any PII detected in traffic to third-party APIs?"

Debugging & Troubleshooting

"What 500 errors occurred on the /api/payments endpoint today?"

"Which endpoints have the slowest average response times?"

Operations & Monitoring

"Compare traffic volume this week vs last week"

"List all unique endpoints my applications connected to this month"

---

Tips for Effective Queries

1. Be specific about time ranges. "In the last hour" or "yesterday" helps narrow results.

2. Use endpoint names. If you know the domain (e.g., api.stripe.com), include it.

3. Start broad, then narrow. Begin with summaries, then drill into specific issues.

4. Ask follow-up questions. The AI maintains context, so you can refine your queries.

---

Enhance with Claude Code Skills

For smarter AI assistance, add a Qpoint skill to your Claude Code configuration. Skills give the AI context about when and how to use Qpoint tools effectively.

Create ~/.claude/skills/qpoint/SKILL.md:

# Qpoint Traffic Analysis

Use the Qpoint MCP tools for API traffic investigation, security audits, and debugging.

## When to Use
- Investigating API traffic patterns and connections
- Security audits (auth failures, PII exposure, shadow IT)
- Debugging (error rates, slow endpoints, failed requests)
- Compliance checks (data exposure, vendor inventory)

## Workflow
1. Start with `traffic_summary` to get an overview of activity
2. Use `list_connections` or `list_requests` with filters to drill down
3. Check `list_issues` for policy violations and errors
4. Use `pii_summary` to audit sensitive data exposure

## Query Tips
- Always specify time ranges for faster queries (e.g., "last hour", "yesterday")
- Use `*Ish` filters (endpointIsh, pathIsh) for fuzzy/substring matching
- Start broad, then narrow with additional filters
- Use `count_*` tools first to understand data volume before listing

## Common Investigations
- **Shadow IT:** "What endpoints are we connecting to that aren't known vendors?"
- **Auth failures:** "Show 401/403 errors in the past week"
- **Performance:** "Which endpoints have the slowest response times?"
- **Data exposure:** "Is there PII in outbound traffic?"

Once saved, Claude Code will automatically use this guidance when working with your Qpoint traffic data.

---

Troubleshooting

"Authentication required" or 401 errors

• Verify your token is correct

• Check the token hasn't expired

• Ensure the Authorization header format is correct: Bearer YOUR_TOKEN

"No data returned"

• Verify your Qtap agents are running and sending data

• Check that the time range includes periods with traffic

---

Next Steps

• MCP Tools Reference — Complete reference for all available query tools

• Qplane Dashboards — Visual exploration of your traffic data