Users in one region cannot reach your service, a customer says your new prefix "is not on the internet yet", or a monitoring alert says someone else is announcing your address space. Answering any of these starts with the same question: what does the global routing table currently say about this prefix, and who is originating it?
The BGP Looking Glass answers that from the outside. Enter an IP address, an IPv4 or IPv6 prefix or an AS number and it shows the origin AS, the AS paths that RIPE RIS route collectors currently hold, BGP visibility, RPKI validation for every observed origin and the origin network's PeeringDB record, plus an interactive AS-path graph and a route comparison view. It is read-only and only reports what the sources observed.
What is a BGP looking glass?
In simple terms
The internet is a network of networks. Each network, called an Autonomous System (AS) and identified by an AS number, tells its neighbors which address blocks (prefixes) it can deliver traffic to. Those announcements travel from network to network using BGP, and each network adds its own AS number to the announcement's AS path. A looking glass lets you peek into a router's routing table and see those announcements: who originates a prefix and through which networks it is reached.
Technical background
BGP-4 (RFC 4271) is the routing protocol between autonomous systems. The terms you will see in the results:
| Term | Meaning |
|---|---|
| Prefix | An address block in CIDR notation, e.g. 1.1.1.0/24. Routers forward traffic using the longest (most specific) matching prefix. |
| Origin AS | The AS at the right end of the AS path: the network that announces the prefix. |
| AS path | The list of ASes the announcement passed through, nearest first. The collector's peer is on the left, the origin on the right. |
| Prepending | An AS repeating its own number (64500 64500 64500) to make a path look longer and less attractive. A traffic-engineering technique, not an error. |
| AS_SET | A set of ASes in curly braces, produced by route aggregation. Deprecated for new deployments (RFC 6472) but still seen. |
| Communities | Tags attached to a route (RFC 1997) that networks use to signal policy. |
| MOAS | Multiple Origin AS: the same prefix announced by more than one origin. Used deliberately for anycast and migrations, but also a symptom of hijacks or leaks. |
| ROA / RPKI | A Route Origin Authorization is a signed statement from the address holder that an AS may originate a prefix up to a maximum length. Route origin validation (RFC 6811) compares routes against ROAs. |
| IRR route object | A route registered in an Internet Routing Registry database; used by many networks to build prefix filters. |
Route collectors. RIPE NCC's Routing Information Service runs route collectors (named RRC00, RRC01 and so on) at internet exchanges and data centers. They hold BGP sessions with many networks, record what those peers announce and never forward traffic. A full-table peer sends its complete routing table, which is why visibility is measured against full-table peers only.
Why use the BGP Looking Glass?
- Confirm a new announcement: after turning up a prefix, check that it is announced, from the right origin AS, and visible to most full-table peers.
- Investigate reachability complaints: see whether the prefix is announced at all, whether a covering less-specific or a more-specific exists, and which upstreams carry it.
- Check RPKI before and after creating ROAs: an Invalid state means networks that enforce route origin validation drop the route.
- Spot unexpected origins: more than one origin AS is flagged, with a reminder that anycast and migrations cause the same picture.
- Review traffic engineering: see where prepending is applied and which ASes sit right before the origin.
- Document a state: CSV, JSON or a printout with sources and fetch times attached.
Who should use it?
Network engineers and NOC staff at ISPs, hosting and cloud providers, peering coordinators checking how their routes propagate, system administrators who want to know who routes an IP address they are troubleshooting, and security analysts triaging a possible hijack or route leak. Developers rarely need it, except when a service is reachable from some networks and not others.
How the BGP Looking Glass works
Your browser sends the query to the NetTools.tech API; the NetTools server queries two fixed public sources and returns one normalized report. No URL or host you type is ever contacted.
Input parsing (in the browser and again on the server). Whitespace is removed. AS13335, 13335 and asdot 1.10 are AS numbers; anything else must be an IPv4 or IPv6 address, optionally with one prefix length. Host bits are cleared (1.1.1.7/24 becomes 1.1.1.0/24, with a note). The tool refuses without any lookup:
- private, documentation, loopback, link-local, CGNAT, multicast, benchmarking and other reserved address space ("… address space is never announced in global BGP routing."),
- IPv4 prefixes shorter than /8 and IPv6 prefixes shorter than /16,
- private-use, documentation and reserved AS numbers (for example AS64500, documentation range of RFC 5398).
Prefix and address lookups. The server calls these RIPEstat data calls:
| Data call | What it provides |
|---|---|
prefix-overview | Whether the resource is announced, origin ASes and holder names, registry block |
looking-glass | The routes RIS collectors currently hold: collector, location, peer IP, AS path, origin, prefix, last update, communities |
routing-status | First and last seen, visibility among RIS full-table peers per address family, origins with IRR route objects, less- and more-specific prefixes |
rpki-validation | RIPEstat's validation result (validator named in the result) for each observed origin, up to six origins, with the matching ROAs |
as-names | AS names for every AS in the paths (decoration only) |
An address is resolved to its covering routed prefix, and the report says so. The primary origin's network record is fetched from PeeringDB (net with its IX connections and facilities) as clearly labelled metadata.
AS number lookups use as-overview, announced-prefixes, asn-neighbours and PeeringDB, and show the same routing overview as the ASN Explorer.
Independent sources. Every call is made separately. If one fails, its line in Sources & freshness shows the error and everything else is still displayed. Each upstream request has a 25-second timeout and one retry on timeouts, HTTP 429 and server errors; the requests are paced per provider so NetTools stays within RIPEstat's and PeeringDB's limits. RIPEstat requests carry a sourceapp identifier as RIPEstat asks clients to do; the optional PeeringDB API key is stored on the server and never sent to the browser.
Caching. Identical concurrent lookups share one upstream request:
| Data | Cache time |
|---|---|
| Route observations (looking glass) | 5 minutes |
| Prefix overview, routing status, RPKI validation | 30 minutes |
| Announced prefixes, AS neighbours | 1 hour |
| AS overview, PeeringDB network | 6 hours |
| AS names | 24 hours |
How to use the NetTools.tech BGP Looking Glass
- Open the BGP Looking Glass.
- Enter an IP address, prefix or AS number, e.g.
8.8.8.8,1.1.1.0/24,2606:4700:4700::1111orAS13335, or click one of the examples. Press Search; progress steps show which source is being queried. - Read the routing summary: the four KPI cards (Origin ASN, BGP visibility, RPKI, Prefix), any notes, and the details list below them.
- Switch between the tabs:
- Observed routes – the route table. Filter with the search box (path, AS name, peer IP, collector) and the Collector, Collector country, Peer ASN, Origin (only when more than one origin is observed) and RPKI selectors; Sort by collector, peer ASN, path length or last updated. 50 routes per page.
- AS path graph – an interactive graph of the distinct observed paths.
- Compare routes – path statistics and a side-by-side view of selected collectors.
- PeeringDB – the primary origin network's published peering information.
- Use the links under the summary to continue in the ASN Explorer, the IXP & Peering Explorer, IP Lookup (for addresses) or WHOIS (registry record of the prefix's address; for the AS holder use the ASN Explorer).
- Export with CSV, JSON, Copy all or Print. The export follows the current view: the filtered route list on Observed routes, the selected collectors on Compare routes.
- Share the URL. It holds the query and the view state:
q,tab,s(search),collector,country,peer,origin,rpki,sortandpage, e.g./bgp?q=1.1.1.0/24&tab=compare.
Tip
Refresh from sources in the Sources & freshness box skips the NetTools cache. Use it right after changing an announcement or a ROA; otherwise cached data may be up to 5 (routes) or 30 (RPKI) minutes old.
Understanding the results
Routing summary
| Result | Meaning | What to look for |
|---|---|---|
| Origin ASN | The observed origin(s); "Origin ASNs (n)" when several | An origin you do not expect |
| BGP visibility | Percentage and "seen by X of Y RIS full-table peers (IPv4/IPv6)" | Low visibility after an announcement; without both numbers it shows collectors with a route and peer sessions instead |
| RPKI | State for the primary origin with a one-line explanation | Any Invalid state |
| Prefix | Routed prefix and "announced in BGP", "not announced" or "announcement state unknown" | An address resolving to an unexpected covering prefix |
| Notes | Host bits cleared, the covering routed prefix, multiple origins, not seen by any collector, RIPEstat messages | Multiple-origin notes deserve a closer look, not panic |
| First seen / Last seen | From RIS history, with prefix and origin | A very recent first-seen for a supposedly old prefix |
| Route objects (IRR) | Route objects per origin | Missing objects can explain filtering by IRR-based networks |
| Less- / more-specific prefixes | Covering and contained prefixes (first 30 more-specifics shown), each a new lookup | An unexpected more-specific attracts traffic away from your prefix |
| RPKI validation | Per origin: state, matching ROAs (prefix, max length, AS, validity), validator | ROA max length shorter than the announced prefix |
The RPKI states are Valid, Invalid (origin AS), Invalid (prefix length), Invalid, Not found and Unknown (shown as ✓, ✕, ○ and ?).
Observed routes
| Result | Meaning | What to look for |
|---|---|---|
| Collector | RIS collector ID, location and scope | The location is where the collector sits, not where your traffic goes |
| Peer | The peer's ASN and IP: the network that sent this route to the collector | Whether a given upstream's customers see you |
| AS path | Chips from peer (left) to origin (right); a repeated AS is collapsed to one chip with ×N; long paths fold with "… N more" | Prepending, AS_SET tag, unexpected ASes |
| Origin, Prefix, RPKI | Origin AS, the exact prefix, its validation state or "not checked" | Routes for a more-specific prefix |
| Last updated | When the collector last received a change for this route (UTC) | Very recent changes in many rows point to churn |
| Graph | Opens the graph with this path highlighted | – |
AS path graph
The graph is drawn only from observed adjacencies: an arrow means the two ASes were next to each other in at least one observed path. Collector peers are on the left, origins on the right, each AS placed by its longest observed distance to an origin; arrow thickness grows with the number of routes. Choose Top 12/30/60/120 paths (12 by default); when one rank has many ASes, for example many collector peers of an anycast origin, that column wraps into several sub-columns so the graph grows sideways. Zoom with −, +, Fit or the mouse wheel, drag to pan. Clicking an AS highlights every path through it and states how many of the observed routes use it; clicking an entry in the "Distinct observed paths" list highlights that path and lists its prepending. Business relationships (customer, provider, peer) are neither shown nor inferred.
Compare routes
The summary lists distinct AS paths, origin ASes, the ASes right before the origin with route counts, path length range with prepending collapsed, routes with prepending (and which ASes prepend) and routes with an AS_SET. Below, tick collectors (the first six are preselected) to see the most common path at each; ASes that differ at the same distance from the origin are highlighted.
Example: refused inputs and reading an AS path
Refused before any lookup. These inputs never reach RIPEstat:
192.0.2.0/24 → Documentation (TEST-NET-1) address space is never announced in global BGP routing.
2001:db8::/32 → Documentation address space is never announced in global BGP routing.
10.1.2.3 → Private (RFC 1918) address space is never announced in global BGP routing.
1.0.0.0/7 → /7 is shorter than any prefix in the global routing table; use /8 or longer.
AS64500 → AS64500 is not used on the public internet (Documentation and examples (RFC 5398)).An address lookup. Searching for the public resolver 1.1.1.1 produces the note that the address is covered by the routed prefix 1.1.1.0/24 and that the routing data refers to that prefix. The Origin ASN card shows AS13335 with its name, and the visibility, RPKI and route counts are whatever RIS and RIPEstat report at that moment, with their fetch times in Sources & freshness.
Reading the paths (illustrative values, documentation ASNs, not real output). Suppose the route table shows three peers:
Peer AS64501 AS64501 → AS64502 → AS64496 ×3
Peer AS64503 AS64503 → AS64502 → AS64496 ×3
Peer AS64504 AS64504 → AS64505 → AS64496- The origin is AS64496 in all three routes, so there is a single origin and no MOAS note.
- AS64496 prepends itself three times towards AS64502 but not towards AS64505. That is deliberate traffic engineering: the origin wants inbound traffic to prefer AS64505.
- Compare routes would list AS64502 (2 routes) and AS64505 (1 route) under "ASes right before the origin", and a path length of 3 ASes with prepending collapsed.
- The graph would draw AS64501 and AS64503 pointing to AS64502, AS64504 pointing to AS64505, and both pointing to AS64496 — and nothing else, because no other adjacency was observed.
Common use cases
- Turn-up check: announced, correct origin, RPKI Valid, high visibility.
- ROA rollout: look up each prefix before and after publishing ROAs and watch for Invalid (prefix length) on more-specifics.
- Hijack or leak triage: an unexpected origin or an unexpected more-specific, then the origin's ASN and PeeringDB data.
- Upstream review: which ASes carry your prefix right before the origin, and whether prepending has the intended effect.
- Support tickets: attach the JSON or CSV export with its source notes.
Troubleshooting
"No RIS route collector currently sees this prefix in BGP"
The prefix is not announced, or is announced only as part of a larger prefix, or it is filtered by the networks that peer with RIS. Check the less-specific prefixes, your router's advertised routes and RPKI. A newly announced prefix can take a few minutes to appear; use Refresh from sources after the cache time.
Low visibility percentage
Some full-table peers do not carry the prefix. Typical causes are an RPKI Invalid state, missing IRR route objects, a prefix longer than /24 (IPv4) or /48 (IPv6) that many networks filter, or an announcement to only one upstream. Compare the RPKI and route-object rows first.
RPKI shows Invalid (prefix length)
A ROA exists for the origin but its maximum length is shorter than the announced prefix, typical when a /22 has a ROA with max length 22 and someone announces a /24 from it. Create a ROA for the more-specific or raise the maximum length deliberately.
Several origin ASes
Read the note: anycast networks and migrations announce the same prefix from several ASes on purpose. Filter Observed routes by Origin to see which collectors see which origin, and check each origin's RPKI state.
"RIPEstat is busy" or "did not answer in time"
NetTools paces its requests and gives up after waiting about 20 seconds for a free slot; upstream requests time out after 25 seconds. The failed part is marked in Sources & freshness while the rest is shown. Try again after a minute.
The result looks stale
Check the fetch time and "(from the NetTools cache)" in Sources & freshness. Route observations may be up to 5 minutes old, RPKI up to 30 minutes; Refresh from sources forces a new query.
Common mistakes
- Treating the collector location as your traffic path. A route at a collector in one city says nothing about where packets from your users go; use Traceroute or MTR for that.
- Reading path differences as an outage. Different peers choose different best paths; that is normal.
- Reading a second origin as a hijack. Multiple origins can be intentional; confirm with the address holder.
- Reading graph arrows as business relationships. They only show adjacency in observed paths.
- Treating PeeringDB as routing evidence. It shows what a network publishes about itself, not what it announces.
- Assuming Not found means insecure or broken. It only means no ROA covers the prefix.
Limitations
- One data source for routes: RIPE RIS via RIPEstat. Other collector projects and operator looking glasses are not queried, so routes that never reach a RIS peer are invisible here.
- Vantage points, not the whole internet: RIS sees what its peers send, which is a sample of networks.
- Visibility depends on RIPEstat's numbers; without both counts no percentage is shown.
- RPKI is RIPEstat's validation; NetTools does not validate ROAs itself, and only up to six origins are checked.
- Cached data (see the table above); not a real-time BGP feed or a monitoring service.
- No relationship inference: the tool does not label customers, providers or peers.
Security & privacy
The BGP Looking Glass is read-only: it never opens a BGP session, never sends announcements or withdrawals, never talks to route servers and never sends traffic to the networks it shows. Your query goes to the NetTools.tech API, which forwards the normalized prefix, address or ASN to RIPEstat and the origin ASN to PeeringDB; your browser contacts those services only if you follow an "open at source" link. Private and reserved space is refused before any lookup, and queries are not stored. The PeeringDB API key stays on the server. There is nothing secret to enter. See the privacy policy.
Related tools
- ASN Explorer – registry holder, announced prefixes, neighbours and PeeringDB for an AS; guide.
- IXP & Peering Explorer – where a network peers, exchange members and facility tenants; guide.
- IP Lookup – owner, network and location data for an address; guide.
- Traceroute and MTR – the actual forwarding path from a probe; traceroute guide, MTR guide.
- WHOIS – registry record for an IP address (such as the prefix's address) or a domain; for AS registry data use the ASN Explorer; guide.
- Network Diagnostics overview.
Summary
The BGP Looking Glass shows how RIPE RIS route collectors currently see a prefix, address or AS: origin, AS paths, visibility, RPKI and the origin's PeeringDB record, each with its source and fetch time. Use it to confirm announcements, check ROAs and triage unexpected origins. Remember that it is a view from RIS vantage points, that path differences are normal and that PeeringDB is self-published metadata.
Frequently asked questions
What is a BGP looking glass?
A looking glass lets you see the BGP routes that a router or route collector holds for a prefix: which AS originates it and which AS paths lead there. The NetTools.tech BGP Looking Glass shows the routes held by RIPE RIS route collectors, read through the RIPEstat Data API.
Where do the AS paths come from?
From the RIPE NCC's Routing Information Service (RIS). Each route is what one network peering with a RIS collector announced to that collector. It is a view from those vantage points, not the path your own traffic takes.
What does the BGP visibility percentage mean?
It is the number of RIS full-table peers that see the prefix divided by the number of full-table peers for that address family, as reported by RIPEstat routing-status. If RIPEstat does not state both numbers, the tool shows the number of collectors with a route instead of a percentage.
What do RPKI Valid, Invalid and Not found mean?
Valid means a ROA authorizes the origin AS for this prefix at this length. Invalid means ROAs exist but none matches the origin AS or the prefix length, so networks that enforce route origin validation drop the route. Not found means no ROA covers the prefix, so RPKI can neither confirm nor reject it.
Why do different collectors show different AS paths?
Every network chooses its own best route, so peers at different collectors often reach the origin through different transit providers. A path difference is normal and is not by itself an outage or a routing problem.
Why does the tool refuse my private or documentation prefix?
Private, documentation, loopback, CGNAT and other reserved address space is never announced in global BGP, so there is nothing to look up and the tool explains this without querying anything. IPv4 prefixes shorter than /8 and IPv6 prefixes shorter than /16 are refused for the same reason.
Is the data real time?
Route observations are cached on the NetTools server for up to 5 minutes, routing status and RPKI for up to 30 minutes and PeeringDB records for up to 6 hours. The Sources & freshness box shows when each part was fetched and the source's own data time, and Refresh from sources bypasses the cache.
Can I use it to announce or withdraw routes?
No. The tool is read-only: it queries public routing databases and never opens a BGP session, sends announcements, talks to route servers or sends traffic to the networks it shows.