Swedish business intelligence
for humans and agents.

The Norric MCP server exposes Swedish economic intelligence — credit risk, procurement signals, BRF data, and company lifecycle signals — through a single Model Context Protocol endpoint. Built for Claude Code, Cursor, Windsurf, and any MCP-compatible agent runtime.

Quickstart

1. Add to Claude Code (~/.claude.json)

{
  "mcpServers": {
    "norric": {
      "type": "http",
      "url": "https://norric-mcp-production.up.railway.app/mcp",
      "headers": {
        "X-Norric-Key": "nk_prod_signal_YOUR_KEY_HERE"
      }
    }
  }
}

2. Call your first tool

# Score a Swedish company for payment risk
curl -X POST https://norric-mcp-production.up.railway.app/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -H "X-Norric-Key: nk_prod_kreditvakt_YOUR_KEY" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/call",
    "params": {
      "name": "kreditvakt_score_company_v1",
      "arguments": {
        "org_nr": "556714-7916",
        "include_graph": false
      }
    }
  }'

3. Get this week's SIGNAL call list

mcp__norric__signal_weekly_call_list_v1(
    vertikal="aldreomsorg",
    limit=10
)

Authentication

Norric uses capability-scoped API keys. Each key grants access to specific product verticals. Keys are passed via the X-Norric-Key header or as a Bearer token.

Key format

nk_{environment}_{capabilities}_{random}

# Examples
nk_prod_signal_a7f3k9x2          # SIGNAL tools only, production
nk_prod_kreditvakt_m2p8q1z4      # Kreditvakt tools only, production
nk_prod_all_b5r7n3w9             # All tools, production
nk_sandbox_all_c4t6y8v2          # All tools, sandbox (free tier)

Response format

Every tool returns the universal NorricResponse envelope. This is non-negotiable — no tool deviates from this schema. Agents can rely on it unconditionally.

SIGNAL tools

Municipality procurement intelligence. 13 signal sources across 290 Swedish municipalities. Operates at procurement step 1 — intent signals — vs Mercell/Opic which operate at step 4.

# Parameters
vertikal: str      # aldreomsorg | skola | it_digital | fastighet | hr | bygg | annat
limit:    int      # 1–50, default 10

# Returns
entries: [{
  rank:            int
  kommunkod:       str         # 4-digit Swedish municipality code
  municipality_name: str
  composite_score: float       # 0–100. Malmö 87.8, Stockholm 85.0
  urgency:         str         # hot | warm | quiet
  components:      list        # 7-component score breakdown
}]

Kreditvakt tools

Swedish B2B payment risk engine. Reads Kronofogden enforcement cases, Skatteverket debt trajectory, and Bolagsverket filings. Backtest validated: 6x lift factor, 9-month median warning, 80% bankruptcy coverage.

org_nr:         OrgnrStr    # Swedish format: XXXXXX-XXXX
include_graph:  bool        # Include ownership chain (default false)

# Returns
signal:         str         # GREEN | AMBER | RED
recommendation_sv: str      # Plain Swedish action e.g. "Kräv förskottsbetalning"
hard_gate_fired: bool
reason_codes:   list[str]
confidence:     float

Roadmap tools

Agent patterns

Norric tools are designed for autonomous agent workflows. Here are the four canonical patterns.

Pattern 1: Due diligence agent

# Before onboarding a new B2B customer
result = mcp__norric__kreditvakt_score_company_v1(org_nr="556714-7916")

if result.data["signal"] == "RED":
    # Agent flags for human review, does not proceed
    escalate(result.data["recommendation_sv"])
elif result.data["signal"] == "AMBER":
    # Agent adjusts payment terms automatically
    set_payment_terms(days=14, upfront=0.5)
else:
    # GREEN — proceed with standard 30-day terms
    proceed_standard()

Pattern 2: Procurement monitoring agent

# Runs nightly — Monday generates the week's call list
call_list = mcp__norric__signal_weekly_call_list_v1(
    vertikal="aldreomsorg",
    limit=10
)

for municipality in call_list.data["entries"]:
    if municipality["urgency"] == "hot":
        briefing = mcp__norric__signal_municipality_briefing_v1(
            kommunkod=municipality["kommunkod"],
            vertikal="aldreomsorg"
        )
        schedule_call(municipality, briefing.data["talking_points"])

Pattern 3: Portfolio monitoring agent (nightly)

# Runs at 02:00 CET — silent unless exception threshold crossed
watchlist = load_customer_org_nrs()  # Your customer portfolio

changes = mcp__norric__norric_watchlist_diff(  # Q3 2026
    org_nr_list=watchlist,
    since_timestamp=yesterday_midnight
)

for entity in changes.data["changed"]:
    if entity["severity"] == "high":
        alert_account_manager(entity)  # Only escalates material changes

TTL and caching

Every NorricResponse includes a ttl field in seconds. Agents must respect this. Do not re-call a tool before the TTL expires — data will not have changed and you will waste call quota.

Pricing — Human tier

Pricing — Agent tier

Pricing — Institutional tier

Pricing roadmap

How Norric's pricing evolves as the AI economy matures. The model shifts from human SaaS → agent-native → data licensing.

Rate limits

Error codes

All errors return a machine-readable error_code alongside a human message. Agents should branch on error_code, not on the message string.

Changelog

Norric AB · MCP Developer Docs · April 2026 · norric.io · edgar@norric.io