Review · VerifiedUpdated 2026-05-03Verified against code 2026-05-03

Verified 2026-05-03 against nudg3-backend/api/public/v1/ (~40 routes counted) and nudg3-mcp-server/src/nudg3_mcp/server.py (10 @mcp.tool() decorators).

Public API & MCP Server

Two programmatic surfaces: the REST API v1 for traditional integrations, and the MCP server for AI tool access.

Public API v1 — Production

Capability	Detail
Analytics endpoints	Visibility, sentiment, competitive position, prompt performance, source analysis
Insights endpoints	Boosters, recommendations (shipped 2026-04-04, commit `c6ee216`)
Export endpoints	CSV/JSON exports for all analytics data
Authentication	API key (SHA-256 hashed), with IP allowlist support
Rate limiting	3 tiers (Starter: 30/min, Professional: 120/min, Enterprise: 600/min), sliding window via Redis
Brute force protection	10 failures/5min = 15min lockout
Audit logging	Workspace-scoped queries, multiple event types
Scope-based access	`read:analytics`, `export:data`, `read:insights`, `write:actions`
Workspace isolation	Mandatory — all requests scoped to the API key’s workspace

Approximately 40 routes across analytics + insights + exports as of 2026-05-03. See Quick Stats for the canonical count and source path.

MCP Server — Production

The MCP (Model Context Protocol) server lets customers query their Nudg3 data from any MCP-compatible AI tool:

Claude Desktop — ask about brand visibility while doing strategy work
Cursor / Windsurf — pull brand data into development and content workflows
Custom agents — integrate Nudg3 intelligence into any AI pipeline

10 tools (verified 2026-05-03)

get_metric_catalog, get_overview, analyze_competitors, analyze_prompts, analyze_sources, analyze_responses, export_data, get_reports, get_insights, get_actions

These are composable investigation tools, not 1:1 endpoint wrappers. The AI client composes queries against a catalog of tools, mirroring how the Ask NUDG3 agent works internally.

Infrastructure

Cloudflare-fronted at api.nudg3.ai, API key authenticated, same rate limiting and workspace isolation as the REST API. Cache-Control headers enable edge caching for read-heavy MCP query patterns.

Enterprise governance angle

MCP is not just a developer feature — it’s an enterprise governance play:

$1.4B–$5.8B AI governance market by 2029-2030 (35-50% CAGR)
69% of organisations cite AI-powered data leaks as top security concern
82% of enterprise marketing teams use AI tools without a governance framework

Nudg3’s MCP server provides governed, workspace-isolated, audit-logged access to brand intelligence through AI tools. For enterprise clients requiring AI council approval, this governance layer is the compliance story that unlocks the pilot.