About IPinfo
IPinfo MCP Server exposes IPinfo address intelligence to AI assistants and other MCP clients. It supports hosted and local deployment, with tools for location, network ownership, privacy indicators, residential proxy detection, and quota checks.
Key features
- 01IP geolocation and metadata lookup
- 02Autonomous system and network details
- 03VPN, proxy, and Tor detection
- 04Residential proxy detection
- 05API quota checks
- 06Hosted HTTP and local stdio modes
Use cases
- 01Determine the location of public IP addresses
- 02Identify the network or organization behind an IP
- 03Check whether an address is a VPN or proxy
- 04Inspect remaining IPinfo API quota
MCP support
The server implements MCP and exposes IPinfo lookup, geolocation, ASN, privacy, residential proxy, and quota tools to compatible clients. It can connect through the hosted Streamable HTTP endpoint or run locally over stdio.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://mcp.ipinfo.io/claude mcp add --transport http ipinfo https://mcp.ipinfo.io/{"mcpServers":{"ipinfo":{"type":"http","url":"https://mcp.ipinfo.io/"}}}Open in Cursor{"servers":{"ipinfo":{"type":"http","url":"https://mcp.ipinfo.io/"}}}Open in VS Code{"mcpServers":{"ipinfo":{"type":"http","url":"https://mcp.ipinfo.io/"}}}{"mcpServers":{"ipinfo":{"serverUrl":"https://mcp.ipinfo.io/"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
6| Parameter | Type | Required | Description |
|---|---|---|---|
| ips | array | Yes | Public IP addresses to query, IPv4 or IPv6, e.g. ["8.8.8.8", "2001:4860:4860::8888"]. Private, loopback, reserved, multicast, and bogon addresses are rejected: they are reported under validation_errors and left out of results. |
| page | integer | No | Page of the result set to return, 1-based. Values below 1 are clamped to 1, and a page past the last one returns no results. |
| detailed | boolean | No | Which IPinfo endpoint to query. Leave false to use the Lite endpoint, which returns country, continent, and basic ASN. Set to true to use the full lookup endpoint, which also returns city-level geolocation, privacy flags (VPN, proxy, Tor, hosting, anycast), and richer AS data; it requires a token whose plan includes that data, otherwise the call fails with ACCESS_DENIED. |
| page_size | integer | No | How many IPs to resolve and return per page. Clamped to a maximum of 1000, the API batch limit. Only the IPs on the requested page are fetched, so a smaller page size consumes less quota per call. |
Details
From the project READMEIPinfo MCP Server
The official IPinfo MCP Server lets AI assistants such as Claude answer questions about IP addresses. Ask where an IP is located, which company or network it belongs to, or whether it's a VPN, proxy, Tor exit node, or residential proxy, and the assistant looks it up with IPinfo data.
It implements the Model Context Protocol (MCP), the open standard AI assistants use to connect to external tools, so it works with any MCP-compatible client. It supports the IPinfo Lite, Core, Plus, and Residential Proxy plans.
For the full guide, see the official documentation.
Installation
You need an IPinfo account. Sign up for free at ipinfo.io/signup.
Hosted server
Point your MCP client at https://mcp.ipinfo.io/ (Streamable HTTP). The first time it connects, it asks you to sign in with your IPinfo account through OAuth.
Local server (PyPI)
The server is published on PyPI as ipinfo-mcp-server and runs over stdio, authenticating with your API token. With uv installed, add it to your MCP client configuration:
{
"mcpServers": {
"ipinfo": {
"command": "uvx",
"args": ["ipinfo-mcp-server"],
"env": {
"IPINFO_TOKEN": "<your-ipinfo-token>"
}
}
}
}
Claude Desktop extension
Download mcp.mcpb from the latest GitHub release and open it with Claude Desktop. You'll be asked for your API token during installation.
MCP Registry
The server is listed in the MCP Registry as io.github.ipinfo/mcp.
The server is also listed on LightNow.
Tools
| Tool | Description | Plan requirement |
|---|---|---|
ipinfo_lookup | Full IP data: geolocation, network, and metadata | Any; detailed: true needs Core/Plus |
ipinfo_geolocate | Geographic location | Any; detailed: true needs Core/Plus |
ipinfo_asn | Autonomous system (network ownership) | Any; detailed: true needs Core/Plus |
ipinfo_check_privacy | VPN, proxy, relay, Tor, hosting, anycast, mobile, satellite flags | Paid plan |
ipinfo_check_residential_proxy | Residential proxy detection | Residential Proxy access |
ipinfo_quota | API usage and remaining quota | Any |
Tools that lack access for the token's plan return an ACCESS_DENIED error.
Common parameters
All tools except ipinfo_quota take a list of IPs and are paginated:
| Parameter | Type | Default | Description |
|---|---|---|---|
ips | string[] | required | Public IPv4 or IPv6 addresses. Private, loopback, reserved, multicast, and bogon addresses are rejected and reported in validation_errors. |
page | integer | 1 | 1-based page of the result set. Values below 1 are clamped to 1. |
page_size | integer | 25 | IPs resolved per page, up to 1000. Only IPs on the requested page are fetched, so smaller pages consume less quota per call. |
ipinfo_lookup, ipinfo_geolocate, and ipinfo_asn also take:
| Parameter | Type | Default | Description |
|---|---|---|---|
detailed | boolean | false | false queries the Lite endpoint. true queries the full lookup endpoint, which returns more fields. |
Common output
IP-based tools return:
| Field | Description |
|---|---|
results | Object keyed by IP with the tool-specific data below. |
errors | Object keyed by IP for IPs the API returned an error for. These IPs are left out of results. |
validation_errors | Object keyed by input for values that aren't valid public IPs. Only present when there are any. |
_pagination | total_results, page, page_size, total_pages, has_next, has_previous. |
_meta | api_calls_made and from_cache. Results are cached in memory, so repeat lookups don't consume API quota. |
If the whole request fails (for example a missing or invalid token), the tool returns an error object instead, with code (ACCESS_DENIED, RATE_LIMITED, INVALID_TOKEN, NO_TOKEN, API_ERROR, or UNKNOWN), message, and suggestion.
ipinfo_lookup
Returns the raw IPinfo API response for each IP.
detailed: false(Lite):ip,asn,as_name,as_domain,country,country_code,continent,continent_code.detailed: true(full lookup):ip,hostname,geo(city, region, country, continent, coordinates, timezone, postal code),as(ASN, name, domain, type),anonymous(proxy, relay, Tor, VPN),mobile, and theis_anonymous,is_anycast,is_hosting,is_mobile,is_satelliteflags. Some fields are only available on Plus.
ipinfo_geolocate
Returns, per IP: ip, country, country_code, continent, continent_code. With detailed: true, also city, region, region_code, latitude, longitude, timezone, postal_code.
ipinfo_asn
Returns, per IP: ip, asn, name, domain. With detailed: true, also type (isp, hosting, business, education) and last_changed.
ipinfo_check_privacy
Returns, per IP: ip, is_anonymous, anonymous (is_proxy, is_relay, is_tor, is_vpn), is_anycast, is_hosting, is_mobile, is_satellite.
ipinfo_check_residential_proxy
Returns, per IP: ip and is_residential_proxy. For residential proxies, also service (proxy service name), last_seen (date), and percent_days_seen.
ipinfo_quota
Takes no parameters. Returns token (omitted when connected with OAuth), requests (day, month, limit, remaining), and per-feature quotas under features.
Configuration
The server is configured through environment variables:
| Variable | Default | Description |
|---|---|---|
IPINFO_TOKEN | API token (stdio transport; over HTTP the token comes from the request) | |
IPINFO_API_BASE_URL | https://api.ipinfo.io | Base URL for api.ipinfo.io endpoints |
IPINFO_LEGACY_BASE_URL | https://ipinfo.io | Base URL for legacy ipinfo.io endpoints (e.g. /me) |
IPINFO_CACHE_TTL | 3600 | Seconds a cached IP result stays fresh |
IPINFO_TRANSPORT | stdio | Transport type (stdio or http) |
HOST | 0.0.0.0 | HTTP host (only for http transport) |
PORT | 8000 | HTTP port (only for http transport) |
IPINFO_OAUTH_ISSUER | OAuth issuer whose access tokens are accepted; OAuth is off when unset | |
IPINFO_MCP_RESOURCE_URL | Public URL of this server, required with IPINFO_OAUTH_ISSUER | |
IPINFO_OAUTH_JWKS_URI | <issuer>/.well-known/jwks.json | Where the issuer publishes its signing keys |
With the http transport, each request authenticates with its own Authorization: Bearer <token> header.
Setting IPINFO_OAUTH_ISSUER turns on OAuth for the http transport. The server then advertises the issuer at /.well-known/oauth-protected-resource, answers requests without a valid credential with 401 Unauthorized, and verifies access tokens against the issuer's signing keys. A verified access token is sent to the IPinfo API in place of an API token. API tokens keep working as before.
Feedback
- Bugs and feature requests: open an issue on GitHub Issues.
- Security vulnerabilities: don't open a public issue; follow the security policy.
- Questions about your IPinfo account or plan: contact [email protected].
Contributing
Contributions are welcome. See CONTRIBUTING.md for the workflow and the requirements a change has to meet.
Development
Prerequisites
- Python 3.14+
- uv
Setup
uv sync --dev
cp .env.example .env
# Add your IPinfo token to .env
Running the server
The server supports two transports: stdio (default) and HTTP.
# stdio (default, used by MCP clients)
uv run ipinfo-mcp-server
# HTTP
IPINFO_TRANSPORT=http HOST=0.0.0.0 PORT=8000 uv run ipinfo-mcp-server
Tests
# All tests
uv run pytest
# Integration tests (requires IPINFO_TOKEN)
uv run pytest tests/integration/
Integration tests hit the real IPinfo API and validate response structure only (no exact value assertions). They require IPINFO_TOKEN to be set and are skipped otherwise.
Type checking
uv run pyright
Linting
uv run ruff check .
uv run ruff format .
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.