Developers · MCP

Connect Navrik to Claude, ChatGPT, and any AI agent.

Navrik is a Model Context Protocol (MCP) server. Point any MCP-compatible client at your workspace and it can read your data, run reports, manage documents, and take action — 166 tools, scoped to an API key you control.

Get an API key REST API reference

Why MCP

One server, every AI client.

166 tools

Your CRM, calendar, documents, analytics, social and billing — exposed as callable tools.

Scoped & secure

Authenticate with a workspace API key; per-tool read/write scopes you set and can revoke.

Any MCP client

Claude, ChatGPT, Cursor, Windsurf — anything that speaks the Model Context Protocol.

Live workspace data

The same service layer as the app, not a copy. Reads — and scoped writes — are real.

Step 1

Create a scoped API key.

In Navrik, open Settings → API keys and create a key with the scopes you want the AI to have (e.g. crm:read, documents:read, proposals:write). The key's scopes decide which of the 166 tools it can call — an agent can be read-only, propose-only, or a full operator.

Open API keys

Step 2

Connect your client.

Add the Navrik MCP server to any client below, pasting your sk_live_ key as the Authorization bearer token.

Claude Code (CLI)

claude mcp add --transport http navrik \
  https://www.navrik.co.uk/api/mcp \
  --header "Authorization: Bearer sk_live_your_key_here"

Claude Desktop

Add to claude_desktop_config.json. The mcp-remote bridge lets Claude Desktop reach a remote server with an auth header.

{
  "mcpServers": {
    "navrik": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote",
        "https://www.navrik.co.uk/api/mcp",
        "--header", "Authorization: Bearer sk_live_your_key_here"
      ]
    }
  }
}

On Claude.ai, add it under Settings → Connectors → Add custom connector and sign in with Navrik when prompted (no key needed). Claude keeps one connector per URL, so to connect more than one Navrik workspace use each workspace’s own URL: https://www.navrik.co.uk/api/mcp/w/<workspace-slug>. That URL is pinned to the workspace and skips the workspace picker at consent.

ChatGPT / OpenAI

Attach Navrik as a remote MCP tool via the Responses API (or add the server URL under ChatGPT’s Connectors in developer mode).

// OpenAI Responses API — attach Navrik as a remote MCP tool
{
  "model": "gpt-5",
  "input": "How many open deals do we have, and what's the total value?",
  "tools": [
    {
      "type": "mcp",
      "server_label": "navrik",
      "server_url": "https://www.navrik.co.uk/api/mcp",
      "headers": { "Authorization": "Bearer sk_live_your_key_here" }
    }
  ]
}

Cursor

// ~/.cursor/mcp.json
{
  "mcpServers": {
    "navrik": {
      "url": "https://www.navrik.co.uk/api/mcp",
      "headers": { "Authorization": "Bearer sk_live_your_key_here" }
    }
  }
}

Any client (raw JSON-RPC)

List the available tools with a single request — handy to verify a key:

curl -X POST https://www.navrik.co.uk/api/mcp \
  -H "Authorization: Bearer sk_live_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Reference

Endpoint & protocol.

Endpoint: https://www.navrik.co.uk/api/mcp (a single POST endpoint)
Transport: MCP Streamable HTTP — POST a JSON-RPC 2.0 message, get a JSON response
Methods: initialize, tools/list, tools/call, ping
Auth: Authorization: Bearer sk_live_… — a workspace API key from Settings → API keys
Scopes: per-tool <domain>:read / <domain>:write, enforced on every call; the key decides which tools are callable
Isolation: the key is bound to one workspace — a client only ever sees that workspace’s data

Bring your workspace to your favourite AI.

Create a scoped key and connect Navrik to Claude, ChatGPT, or your own agent in minutes.

No pressure. Tell us what is taking too much time, getting missed or holding back growth.

Talk to Navrik