MCP Server

​

What it is

• A bridge from your assistant to TigerGate. The findings, Issues, assets, and remediation flows you see in the dashboard become tools the model can call.
• Read-first. Search and inspect findings, correlated Issues, asset inventory, exposure graphs, and CVE intel. All read-only.
• Remediation when you want it. Preview an auto-fix (a diff, no PR), open a PR, file a ticket, or update a finding’s status — each behind an explicit write gate.
• Skills that compose the primitives. Higher-level playbooks like “triage my top risks” or “remediate this finding end-to-end” chain the individual tools for you.

​

Prerequisites

​

Connect

​

Find your region first

claude mcp add --transport http tigergate https://mcp.tigergate.dev/mcp \
  --header "Authorization: Bearer tg_..."

{
  "mcpServers": {
    "tigergate": {
      "type": "http",
      "url": "https://mcp.tigergate.dev/mcp",
      "headers": { "Authorization": "Bearer tg_..." }
    }
  }
}

{
  "mcpServers": {
    "tigergate": {
      "url": "https://mcp.tigergate.dev/mcp",
      "headers": { "Authorization": "Bearer tg_..." }
    }
  }
}

{
  "servers": {
    "tigergate": {
      "type": "http",
      "url": "https://mcp.tigergate.dev/mcp",
      "headers": { "Authorization": "Bearer tg_..." }
    }
  }
}

{
  "mcpServers": {
    "tigergate": {
      "serverUrl": "https://mcp.tigergate.dev/mcp",
      "headers": { "Authorization": "Bearer tg_..." }
    }
  }
}

experimental_use_rmcp_client = true

[mcp_servers.tigergate]
url = "https://mcp.tigergate.dev/mcp"
http_headers = { Authorization = "Bearer tg_..." }

{
  "mcpServers": {
    "tigergate": {
      "command": "npx",
      "args": ["mcp-remote", "https://mcp.tigergate.dev/mcp", "--header", "Authorization:${AUTH}"],
      "env": { "AUTH": "Bearer tg_..." }
    }
  }
}

​

The tool catalog

{
  "id": "…",
  "product": "cloud | code | container | pentest | kspm | runtime",
  "finding_type": "sca | sast | secret | iac | …",
  "severity": "critical | high | medium | low | info",
  "status": "open | in_progress | resolved | false_positive | accepted | wont_fix | muted | fixed",
  "title": "…",
  "resource": { "type": "…", "name": "…", "uid": "…", "region": "…", "account": "…", "repository": "…", "file": "…", "line": 0 },
  "cve_id": "CVE-…",
  "cvss": 0.0,
  "epss": 0.0,
  "kev": false,
  "remediation": { "text": "…", "fixed_version": "…", "url": "…", "auto_fix_available": true },
  "first_seen": "…",
  "last_seen": "…"
}

​

Read tools (read-only)

​

Remediation tools (write — gated)

​

Skills (higher-level playbooks)

​

Security and the MCP firewall

• Read-only database role. Read tools query Postgres/ClickHouse through a role that has no write grants. A read tool physically cannot mutate your data.
• Write gate. Write tools are only registered when writes are enabled server-side, and each destructive apply requires human approval in your client before it runs. Nothing opens a PR or changes a status silently.
• Delegated writes. Writes never touch the database directly — they go through the backend’s normal, audited code path, so an MCP-initiated change behaves exactly like a dashboard-initiated one.
• Prompt-injection neutralization. Findings and Issues contain attacker-controlled text (a commit message, a resource name, a page title). The firewall neutralizes injected instructions in results before they reach the model, so a malicious finding can’t hijack your assistant.
• Secret redaction. If a finding’s evidence contains a secret, it’s redacted on the way out — the assistant sees that a secret exists, not its value.
• Org scoping. The API key binds every call to one organization. There is no cross-org tool and no way to widen scope from a request.
• Allow / deny + rate limiting. Tools can be allow-listed or denied, and calls are rate-limited per key.
• Audit. Every tool call is recorded — tool, parameters, key, and outcome — the same way API key usage is audited.

​

Example: triage and fix in one conversation

You: List my critical cloud findings.

{
  "tool": "tigergate_findings_search",
  "arguments": { "product": "cloud", "severity": "critical", "status": "open", "limit": 10 }
}

You: What's my highest-risk open code vulnerability, and can you fix it?

{ "tool": "tigergate_findings_search",
  "arguments": { "product": "code", "finding_type": "sca", "severity": "critical", "status": "open", "limit": 5 } }

{ "tool": "tigergate_get_remediation",
  "arguments": { "product": "code", "finding_type": "sca", "id": "<finding-id>" } }

{ "tool": "tigergate_fix_preview",
  "arguments": { "finding_type": "sca", "id": "<finding-id>" } }

{ "tool": "tigergate_fix_apply",
  "arguments": { "ids": ["<finding-id>"] } }

{ "tool": "tigergate_fix_status", "arguments": { "attempt_id": "<attempt-id>" } }

​

FAQ