Model Context Protocol · Norway
Mynk MCP Server documentation
Mynk exposes a remote Model Context Protocol (MCP) server at https://mcp.mynk.no so AI tools — claude.ai, Claude Code, Cursor, VS Code — can search 500K+ Norwegian companies and people, and reveal verified contact details, directly from a chat session. This guide covers everything you need: connecting your client, the full tool reference, credits and troubleshooting.
Requires a Mynk workspace on the Pro plan. Transport is streamable HTTP — any MCP client that supports remote HTTP servers can connect.
Connect with claude.ai
claude.ai connects through OAuth — no API key needed. You add Mynk as a custom connector once, approve access in your browser, and every chat can then use the Mynk tools.
In claude.ai, go to Settings → Connectors and click Add custom connector.
Name the connector Mynk and enter the URL
https://mcp.mynk.noLeave the OAuth Client ID and Client Secret fields empty — claude.ai registers itself automatically — and click Add.
claude.ai opens the Mynk consent page in your browser. Log in with your normal Mynk account.
Review the permissions and approve. You're sent back to claude.ai, and the Mynk tools are ready to use in your chats.
Access requires a Pro-plan workspace. Login and approval happen on Mynk's own consent page — claude.ai never sees your Mynk password.
Connect with an API key
Header-capable MCP clients — Claude Code, Cursor, VS Code and most others — authenticate with a static API key sent as a bearer token on every request.
1. Create an API key
In the Mynk dashboard, go to Settings → API Keys, name the key (for example “Claude Code on my laptop”) and click Create Key. The full key (mynk_...) is shown exactly once — copy it immediately; Mynk cannot show it again. You can have up to 5 active keys, and revoking one stops it working within seconds.
2. Claude Code
claude mcp add --transport http mynk https://mcp.mynk.no \
--header "Authorization: Bearer mynk_..."3. Other MCP clients (Cursor, VS Code, etc.)
Point the client at the same URL and header — the transport is streamable HTTP:
URL: https://mcp.mynk.no
Transport: streamable HTTP
Header: Authorization: Bearer mynk_...Available tools
The server exposes eight tools for Norwegian B2B sales intelligence: company search, exhaustive deep search, company profiles, employee and decision-maker lookups, people search, and credit-gated contact reveal.
vat_number is the Norwegian organisasjonsnummer (org number) — it ties everything together: company search returns it, and company profiles and employee lookups take it as input. Person results include an id you pass on to get_person_profile and reveal_contact.
| Tool | What it does |
|---|---|
| search_companies | Fast natural-language search across Norwegian companies. Returns a vat_number (org number) per company, and total_matched showing how many companies matched in total. ~15–30s fresh, instant when cached. |
| start_deep_search | Exhaustive search covering every match, not just the top candidates — same coverage as the Mynk web app. Runs as a background job and returns a job_id. |
| get_search_results | Poll a deep-search job and page through its complete result set, 25 at a time. |
| get_company_profile | Full company profile by org number: registry data plus AI-enriched fields (summary, is_b2b/is_b2c, who they serve, what they sell, ideal client profile). |
| get_company_employees | List known employees and roles at a company, with an optional natural-language role filter like “decision makers” or “CTO”. |
| search_people | Search people across all Norwegian companies by keywords, location, or current employer. |
| get_person_profile | A person's profile by ID: role, seniority, department, employer, LinkedIn. |
| reveal_contact | Reveals a person's verified work email and phone number. Spends workspace credits — a separate, explicit tool so a model never invokes it speculatively. |
Fast vs deep search
Two search modes with an explicit tradeoff between speed and coverage. Start fast; when a broad query reports a large total_matched and you want everything, escalate to a deep search.
search_companies (fast)
- Answers in ~15–30 seconds — instantly when the query is cached.
- On broad queries, covers the ~150 candidates most relevant to your query, ranked semantically.
- total_matched always shows the full match count, so you know what fast mode is leaving out.
- Results come back one page per call, up to 25 companies at a time.
start_deep_search (deep)
- Covers ALL matches — identical coverage to the Mynk web app.
- Runs as a background job, typically 30 seconds to 3 minutes; poll with get_search_results.
- Once completed, the full result set is paged 25 at a time — thousands of results are fine.
- Use it when the fast search reports a large total_matched and you need the complete list.
Example prompts
Once connected, you talk to Mynk in plain language inside your normal chat — no query syntax to learn.
- “Find accounting firms in Bergen with more than 10 employees”
- “Search for IT security companies in Oslo and show me their revenue and employee count”
- “Who are the decision makers at [company name]?”
- “Find the CFO at every SaaS company in Trondheim and get their contact details”
- “Give me every marketing leader in Norwegian companies located in Bergen”
- “Do a full search for all IT services companies in Norway, not just the top matches”
A typical tool flow
Under the hood, a session usually chains the tools like this:
search_companies("accounting firms in Bergen with more than 10 employees")
→ pick a company, grab its vat_number
get_company_profile(vat_number)
get_company_employees(vat_number, role="decision makers")
→ pick a person, grab their id
reveal_contact(person_id) # only when you explicitly want contact infoFor broad queries where you need every match, the model escalates to the deep-search job:
start_deep_search("IT services companies") → job_id
get_search_results(job_id) → status: running
... wait 20-30s, poll again ...
get_search_results(job_id) → status: completed, total: 3570, first 25
get_search_results(job_id, offset=25) → next pageCredits & limits
Searching is unlimited on Pro. The only tool that spends workspace credits is reveal_contact — and it's deliberately a separate tool, so contact details are only ever revealed when you explicitly ask for them.
| Reveal | Credits |
|---|---|
| LinkedIn profile | 1 |
| Verified work email | 1.5 |
| Direct phone number | 10 |
| Full reveal (maximum per person) | 12.5 |
- Already-revealed contacts are returned free — you're never charged twice for the same person.
- Company and people searches are unlimited on the Pro plan; cached repeats are always free.
- Requests are rate-limited to 60 per minute per credential to keep the service stable for everyone.
- Deep-search results are paged 25 companies at a time — ask for the next page to continue.
Security
Every request must carry a valid bearer credential — an API key or OAuth token — tied to an active Pro-plan workspace.
claude.ai connects via OAuth 2.1 with mandatory PKCE. No client secrets are involved, and authorization codes are single-use and short-lived.
Mynk never stores your raw credentials: API keys and OAuth tokens are stored as SHA-256 hashes only. A key is shown once, at creation.
OAuth access tokens are short-lived, and refresh tokens rotate on every use.
Contact details are only revealed through the explicit reveal_contact tool — search results never include unrevealed emails or phone numbers.
Revoking a key or downgrading a plan cuts access within seconds.
Troubleshooting
All errors come back as clean, readable messages in the chat — these are the ones you're most likely to see.
401 Unauthorized
The credential is missing, invalid, expired, or revoked.
Check the Authorization header in your client config, or create a fresh API key under Settings → API Keys. For claude.ai, remove and re-add the connector to re-run the OAuth flow.
403 Forbidden
Your workspace is not on the Pro plan.
MCP access is a Pro feature — upgrade in the Mynk dashboard or see the pricing page.
429 Too Many Requests
More than 60 requests in one minute on the same credential.
Wait the number of seconds given in the Retry-After header — the model usually handles this by itself.
Search timeout
A very broad fast search hit its time budget.
Retry, or use start_deep_search — the background job has no timeout pressure and covers every match.
Insufficient credits
reveal_contact needs more credits than the workspace has left.
The error message shows how many credits are required versus available in your workspace.
MCP Server FAQ
- A remote Model Context Protocol server at https://mcp.mynk.no that lets AI tools like Claude search Norway's company and people database directly inside a chat session — no separate app, no copy-pasting between tools.
- Any MCP client that supports streamable HTTP: claude.ai (via a custom connector), Claude Code, Cursor, VS Code, and others. ChatGPT support is not confirmed today.
- Yes. MCP access is a Pro-plan feature, gated the same way as the rest of Mynk's API access. See the pricing page for current rates.
- Searches are unlimited on Pro and cached searches are free. Only reveal_contact spends credits, and already-revealed contacts are returned free.
- No. Contact details (email/phone) are only revealed when reveal_contact is explicitly called, and already-revealed contacts are returned free without a second charge. Search results are trimmed to a public field whitelist before they ever reach the model.