Platform · For developers

Connect your AI agent to NorthBound Sentinel

The NorthBound Sentinel MCP server is a remote server for the Model Context Protocol (MCP). Connect your AI coding agent and the assistant that writes your code can also read your attack surface and findings, and start an assessment on a domain you own. It is open to every plan and metered by credits.

01 Capabilities

What your agent can do

Exact tool names and parameters come from the server itself (tools/list), so your client always shows the current set. In plain terms:

  • See your inventory

    List your projects and domains, with each domain’s ownership verification status and open-finding counts.

  • See your exposed surface

    For a domain, retrieve the application map the platform discovered, its endpoints, and the technologies it runs together with known-vulnerability alerts.

  • Read findings

    Search and filter findings by project, domain, severity, status or keyword, and open one to get its evidence, proof level, severity scoring and remediation guidance.

  • Get a posture summary

    Open findings by severity, how much of your inventory is ownership-verified, remediation deadlines that are slipping, and the top things to fix next.

  • Follow and launch assessments

    List past and running assessments, poll the status of one, or launch an assessment on a domain you own, with the safeguards below.

  • Generate reports

    Produce a report for a project or a single domain as a structured summary, a standalone HTML document or a digitally signed PDF.

02 Connect

Connect in three steps

Use it from Claude Code, Cursor, VS Code or any other MCP client that supports remote HTTP servers with custom headers.

  1. Create an API key

    In the app, open Settings > API Credentials and create a credential. Pick only the scopes you need and, if you like, restrict the key to an IP allowlist. The key starts with nbk_live_ and is shown once, so copy it right away. If you lose it, rotate it or create a new one.

  2. Add the server to your client

    Add the server to Claude Code, Cursor, VS Code or any other MCP client (setup below). Use the server URL and send the key in the Authorization header: the word Bearer, a space, then the key. Keep the key in an environment variable or a secret store, never in a repository.

  3. Ask your agent

    For example: “List my NorthBound assets and tell me which ones have open high or critical findings.” Or: “Show me the evidence and the fix guidance for that finding, then propose a patch.” Or: “Start a passive assessment on staging.acme.example and tell me when it finishes.”

03 Server details

Server details

Server details
ItemValue
Server URLhttps://app.northboundintelligence.com/api/mcp
TransportStreamable HTTP: JSON-RPC 2.0 over POST, with JSON responses
AuthenticationAPI key in the header Authorization: Bearer nbk_live_.... The server also accepts the key in an X-API-Token header.
HandshakeThe handshake (initialize and tools/list) is public, so a client can see what is available. Calling a tool needs a valid key.
Sign-inAPI key only. OAuth sign-in is not supported yet.
Protocol versions2025-03-26 and 2024-11-05
04 Setup

Set up your client

Put the key in an environment variable named NORTHBOUND_API_KEY, then add the server to your client. The header value is the word Bearer, a space, then the key. Key names differ slightly between clients; these are the working configurations for the three most common ones.

Set your key in the environment

export NORTHBOUND_API_KEY="nbk_live_..."

Claude Code: one command

claude mcp add --transport http northbound-sentinel \
  https://app.northboundintelligence.com/api/mcp \
  --header "Authorization: Bearer $NORTHBOUND_API_KEY"

Claude Code: .mcp.json at the project root

{
  "mcpServers": {
    "northbound-sentinel": {
      "type": "http",
      "url": "https://app.northboundintelligence.com/api/mcp",
      "headers": {
        "Authorization": "Bearer ${NORTHBOUND_API_KEY}"
      }
    }
  }
}

Cursor: ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (this project)

{
  "mcpServers": {
    "northbound-sentinel": {
      "url": "https://app.northboundintelligence.com/api/mcp",
      "headers": {
        "Authorization": "Bearer ${env:NORTHBOUND_API_KEY}"
      }
    }
  }
}

VS Code: .vscode/mcp.json

{
  "inputs": [
    {
      "type": "promptString",
      "id": "northbound-api-key",
      "description": "NorthBound API key (nbk_live_...)",
      "password": true
    }
  ],
  "servers": {
    "northbound-sentinel": {
      "type": "http",
      "url": "https://app.northboundintelligence.com/api/mcp",
      "headers": {
        "Authorization": "Bearer ${input:northbound-api-key}"
      }
    }
  }
}

List the tools from a terminal

curl -sS https://app.northboundintelligence.com/api/mcp \
  -H "Authorization: Bearer $NORTHBOUND_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Claude Code: the one-command option stores the expanded key in your local Claude Code configuration, while the .mcp.json option holds only a reference, so it is safe to commit. Cursor resolves ${env:NAME} and needs the variable visible to the Cursor process; restart Cursor after changing it. VS Code uses servers as its root key, not mcpServers, and can prompt for the key so it is never written to the file. Then confirm that the server shows as connected: claude mcp list (or /mcp inside Claude Code), the MCP section in Cursor’s settings, or MCP: List Servers in VS Code.

05 Permissions

Scopes: give the key only what it needs

Scopes: give the key only what it needs
ScopeWhat it lets the agent do
read_assetsRead your projects, assets, discovered surface and assessment runs
read_findingsRead findings and your posture summary, and generate reports
launch_scansStart assessments. Leave it off if you only want read access.

A read-only key cannot start an assessment.

06 Plans and credits

On every plan, metered by credits

MCP access is not reserved for a particular plan. Every plan includes a monthly allowance of API credits, and tool calls use them.

On every plan, metered by credits
PlanAPI credits per month
Free50
Free trial500
Pro500
Business1,000
Elite2,000
CorporateScoped to your organization

If a call is not covered by your plan or your remaining allowance, the server answers 402. See the plans.

07 Safeguards

Safeguards built into the server

Your agent works inside the same rules as the rest of the platform.

  • Your data only

    Every call is limited to your own account. Results never include another customer’s data.

  • Least privilege

    Keys carry scopes. A read-only key cannot start an assessment.

  • Ownership first

    Anything more active than passive observation requires verified ownership of the domain. Unverified domains accept passive assessments only.

  • A person confirms authorization

    Launching an assessment requires an explicit confirmation that you are legally authorized to test the target, recorded in the audit trail. Your agent must ask you; it must not confirm on its own.

  • One at a time, rate limited

    Assessments run one after another per user, one launch per request, and calls are rate limited per key.

  • Client-safe output

    Results use the same client-facing wording as the NorthBound app and reports.

08 Troubleshooting

If a call fails

If a call fails
SymptomLikely cause
401The key is missing, mistyped, expired or revoked; the request comes from an IP address outside the key’s allowlist; or the header is not Authorization: Bearer <key>.
402Your current plan or usage allowance does not cover the call. Check your plan and credits in the app.
Insufficient scopeThe key lacks the scope that tool needs. Create or edit a key with it.
429 or rate limitedToo many calls or launches in a short time. Wait and retry.
Variable name sent literallyThe environment variable is not set in the process that launched your client. Claude Code warns and sends the literal text, which the server rejects with 401.
09 What it is not

What this is, and what it is not

NorthBound is a human-led, AI-assisted security team. This server gives your agent governed access to your own results and to assessments you are authorized to run. It supports your review; it does not replace it, and it is not a compliance certification or a statement that a system is secure.

Assessments run only on domains you own or are authorized to test. To report a vulnerability in NorthBound itself, see how to report it.

Create a key and connect your agent.

Log in to the app, open Settings, then API Credentials. Questions about access or plans? Write to contact@northboundintelligence.com.

Facts on this page verified 11 November 2026.