routeros-bgp
Configuring BGP on MikroTik RouterOS 7. Covers instances, connections, templates and sessions, eBGP and iBGP, route reflectors and dynamic peers, multihop and BGP unnumbered, advertising networks with address lists, input and output filter chains and accept-nlri, communities, next-hop handling, RFC 9234 roles for route-leak prevention, RPKI origin validation, BFD, monitoring advertisements, and EVPN and L3VPN pointers. Use when peering with a provider or IXP, building an iBGP full mesh or route-reflector setup, filtering or advertising prefixes, validating origins, or when a session will not establish or routes are missing.
How do I install this agent skill?
npx skills add https://github.com/wiresage/routeros-skills --skill routeros-bgpIs this agent skill safe to install?
- Gen Agent Trust Hubpass
This skill provides comprehensive technical documentation and configuration examples for Border Gateway Protocol (BGP) on MikroTik RouterOS 7. The content consists entirely of informational markdown and RouterOS CLI snippets, with no executable scripts or security risks identified.
- Socketpass
No alerts
- Snykpass
Risk: LOW · No issues
What does this agent skill do?
RouterOS BGP
BGP in v7 is redesigned around four menus:
| Menu | Role |
|---|---|
/routing/bgp/instance | The BGP process: as and router-id. Every connection belongs to one instance, and best-path selection only compares routes received by the same instance |
/routing/bgp/template | Protocol parameters (AS, input and output sections, filters). Inherited by connections (templates=) and used for dynamic peers. The built-in default template needs your own AS if you use it: /routing/bgp/template/set default as=65531 |
/routing/bgp/connection | One peer (or a listening subnet): instance, remote.address, local.role, connect, listen, plus any template parameter |
/routing/bgp/session | Read-only state of established, pending and recently dropped sessions (cache survives a flap to avoid route churn) |
Template parameters are also exposed on the connection, so a template is optional for simple setups. Parameters are grouped by direction:
input.*, output.*, local.*, remote.*. The global router ID comes from /routing/id (default: highest interface address).
A first session
# Router1, AS 65531
/routing/bgp/instance/add name=inst1 as=65531
/routing/bgp/connection/add name=toR2 instance=inst1 remote.address=192.0.2.2 local.role=ebgp output.network=bgp-nets
# Router2, AS 65532
/routing/bgp/instance/add name=inst1 as=65532
/routing/bgp/connection/add name=toR1 instance=inst1 remote.address=192.0.2.1 local.role=ebgp output.network=bgp-nets
/routing/bgp/session/print
local.roleis mandatory:ebgp,ibgp,ibgp-rr, or an RFC 9234 role (below). The remote AS is learned from the OPEN message; setremote.asonly to accept a specific AS.- eBGP to a peer that is not on a connected network fails unless
multihop=yes; then setlocal.address(the oldupdate-source). Without it BGP picks an address (highest loopback, else highest address; for a connected eBGP peer the address on that network). - iBGP is multihop by default and normally runs between loopbacks over an IGP (
routeros-ospf). A route learned from an iBGP peer is not passed to another iBGP peer (split horizon): use a full mesh or a route reflector (local.role=ibgp-rron the reflector; clients peer only with it; ORIGINATOR_ID and CLUSTER_LIST prevent loops). input.allow-asrelaxes AS-path loop prevention (needed for some MPLS L3VPN CE setups);nexthop-choice=propagatekeeps the original next hop (EVPN overlays).- Listening on a subnet (
listen=yeswith a prefix asremote.address, for dynamic peers) exposes a DoS surface; protect it with the firewall. - BGP unnumbered: empty
remote.addressandlocal.addressset to an interface; one connection per interface, IPv6 link-local discovery. - Sessions that never start leave no session entry: a TCP block (firewall, MD5 mismatch) shows nothing in
/routing/bgp/session. Check the log (/log/print where topics~"bgp") and that TCP 179 is accepted ininput.
Advertising networks
There is no network menu: put prefixes in an address list, make sure the prefix exists in the routing table (a blackhole route is the
usual anchor), and reference the list from the connection. BGP only advertises prefixes that are present in the routing table; synchronisation
with IGP routes cannot be turned off.
/ip/firewall/address-list/add list=bgp-nets address=198.51.100.0/24
/ip/route/add dst-address=198.51.100.0/24 blackhole
/routing/bgp/connection/set toR2 output.network=bgp-nets # or output.network-blackhole=yes to auto-create the route
Filters
Filters are routing filter chains (routeros-routing). The default action of a chain is reject. If no output chain is set, BGP
advertises every active route it can find: always set output.filter-chain on external sessions.
/routing/filter/rule/add chain=bgp_in rule="if (dst in 192.168.0.0/16) { accept }"
/routing/filter/rule/add chain=bgp_in rule="set bgp-local-pref -1; accept"
/routing/bgp/connection/set toR2 input.filter=bgp_in output.filter-chain=bgp_out
input.accept-nlri=<address-list> (and other input.accept-*) drops unwanted prefixes before they are parsed and stored, saving memory;
a normal input filter only rejects, so the route stays visible as "not active, filtered". Use bgp-med, bgp-local-pref,
bgp-path-prepend, communities and large communities in rules; bgp-out-med belongs in the eBGP output chain.
Route leaks and RPKI
- RFC 9234 roles protect transit relationships: use
ebgp-provider/ebgp-customer,ebgp-rs/ebgp-rs-client,ebgp-peer. The router sends the Role capability in OPEN and ends the session with Role Mismatch (OPEN error subcode 8) if the pair is wrong.ebgp,ibgpandibgp-rrdo not negotiate roles. - RPKI:
/routing/rpki/add group=rpki address=192.0.2.10 port=8282connects to an RTR validator; then in a filterrpki-verify rpkifollowed byif (rpki invalid) { reject } else { accept }. States:valid,invalid,unknown,unverified. Test one prefix with/routing/rpki/rpki-check group=rpki prfx=... origin-as=....
Operation and troubleshooting
/routing/bgp/session/print(flagE= established,.capabilities,.messages,last-notification).- Output advertisements are not recorded by default. Set
output.keep-sent-attributes=yes, then/routing/bgp/session/dump-saved-advertisements 0 save-to=out.pcap. - Per-process load:
/routing/stats/process/print interval=1. BGP input and output can be pinned withinput.affinityandoutput.affinity(alone,afi,instance,vrf,remote-as,main). - Big tables need memory: check
/system/resource/printand useaccept-nlriand filters. Long-lived sessions across BFD:use-bfd=yeson the connection. - Limits:
input.limit-process-routes-ipv4and-ipv6protect the router from a peer that sends too much.
Where to look
| Need | Open |
|---|---|
| Session types, local address selection, unnumbered, dynamic peers | references/guides/user-guides/routing-and-networking-protocols/unicast/bgp/peering-sessions.md |
| Protocol basics, attributes, best-path selection | .../unicast/bgp/understanding-bgp.md |
| Next-hop handling | .../unicast/bgp/nexthop-selection.md |
| Roles and route-leak prevention | .../unicast/bgp/route-leak-prevention.md |
| RPKI | .../unicast/rpki.md |
| EVPN (L2 and L3 overlays) | .../unicast/evpn.md |
| FAQ | .../unicast/bgp/faq.md |
Properties of /routing/bgp/* and /routing/rpki/* | references/cli/routing/bgp/, references/cli/routing/rpki/ |
| Everything else | references/INDEX.md |
Related skills: routeros-routing (filters, tables, BFD, v6 migration), routeros-ospf (IGP for iBGP), routeros-mpls
(L3VPN, VPNv4), routeros-firewall (TCP 179), routeros-ipsec (protecting sessions).
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/wiresage/routeros-skills/routeros-bgp">View routeros-bgp on skillZs</a>