Settings
Settings covers the proxy listener, upstream proxy, TLS fingerprint, system capture, appearance, language, keyboard shortcuts, updates and Lockdown Mode. Appearance, shortcuts and other machine-level choices are not part of a project export.
Finding a setting
Use the search box at the top of Settings to search across every tab, including ones you have not opened. English and Korean names, option labels and related keywords are searchable. Select a result to open its tab, scroll to the section and briefly highlight it. Use the arrow keys and Enter to choose a result, or Escape to dismiss results. With reduced motion enabled, the destination gets a steady highlight instead of a pulse.
Doctor certificate checks
On Windows, Doctor compares the current proxy CA with the local Windows Trusted Root Certification Authorities store for HTTPS use. Installing an old CA with the same name is not enough. If it is missing, install the certificate from mitm.it into that store for the current user or local computer and run Doctor again. Store inspection failures are shown as Not verified, rather than as a certificate error. Browsers with separate certificate stores and other devices still need their own trust setup. The check makes no test connection and does not change TLS verification.
Proxy listeners and API port
Port 8080 is a popular default and another tool may already have it. Open Settings > Proxy listener to edit the primary listener or add more listeners. Each listener can use a different bind address and port, so a local browser, phone, virtual machine, and testing tool can connect to the same project through separate endpoints. All listeners share the project's capture history, CA, upstream route, TLS settings, scope, and Lockdown policy.
One listener is marked Primary. Lanius uses it for the built-in browser, Dashboard address, and setup instructions. Changing the primary does not stop the other listeners. If one saved port is unavailable at launch, the other listeners still start and the failed listener shows its own error. A listener change is saved only after every configured address binds successfully; on failure Lanius restores the previous collection.
By default the proxy accepts connections from this machine only. Choose All interfaces or a specific address to let a phone or a virtual machine point at it, then set their proxy to that listener's address and port. Anyone who can reach that address can send traffic through your proxy, so prefer a specific address over all interfaces on a network you do not control.
The API and MCP port is described in ai-agents.md.
Upstream proxy
To send browser traffic through other proxies, open Settings > Proxy > Upstream proxy, choose Use upstream proxy, and enter HTTP, HTTPS, or SOCKS5 proxy URLs such as http://127.0.0.1:8081 and socks5://127.0.0.1:1080. Add and reorder hops to choose the path from Lanius to the destination. The browser still connects to Lanius. The chain is saved with the project and takes effect when applied.
Enable Use authentication on any hop to set its username and password. After saving, leave the password field blank to keep the existing password. Credentials are stored in the project database, omitted from shareable JSON exports, and included in complete SQLite backups. HTTPS proxy certificates are verified. SOCKS5 routing covers TCP connections; UDP relay is not part of the browser listener. Multi-hop, authenticated, and SOCKS5 routing use a bundled GOST bridge on a loopback port.
Upstream HTTP version
Settings > Proxy > Upstream HTTP version sets the project default for requests Lanius sends to target servers. Choose Auto, Force HTTP/1.0, Force HTTP/1.1, or Force HTTP/2. This controls the origin connection, independently of the browser-to-Lanius protocol and the upstream proxy's CONNECT connection. Product update checks and other control requests are unaffected.
Auto preserves browser negotiation; standalone Replay/Fuzzer sends offer h2 and HTTP/1.1 over HTTPS. HTTP/1.0 uses a Content-Length body and closes the connection. Forced HTTP/2 requires HTTPS and successful h2 negotiation; it never silently falls back to HTTP/1.1. Plaintext h2c is unsupported. WebSocket upgrades require Auto or HTTP/1.1. Forced modes buffer streamed uploads before applying request version and body framing changes. Plugins using callable request stream transforms must use Auto.
Changing the default closes proxy connections and stops active Replay/Fuzzer sends. The choice is saved in the project database and included in exports. Replay tabs and Fuzzer drafts can inherit it or override it for a request/run. A Fuzzer run records the effective mode when started. Scope and Lockdown checks still apply before DNS or socket creation.
History details and Replay results display the actual server direction protocol. Older captures have no recorded upstream protocol; their request-line version alone does not establish how Lanius connected to the server.
Observed public IP
Settings > Proxy > Observed public IP can keep an audit trail of the public address seen outside this machine. It is disabled by default. When enabled, Lanius checks IPv4 and IPv6 separately every one, three, or five minutes. The default observers are https://api.ipify.org for IPv4 and https://api6.ipify.org for IPv6; each can be replaced with a custom HTTPS endpoint. A custom endpoint may return a plain IP address or JSON containing an ip, ip_addr, or address field.
The first successful observation and each address change create a project database row. An unchanged address only advances that row's last-seen time and check count, so routine polling does not fill the database with duplicate rows. The audit trail is included in project exports and SQLite backups. These requests follow the configured upstream route but are not added to Proxy History. An observed public IP proves what the observer saw; another target may still see a different route or address.
Use View IP history to inspect each address period, including its first and last observation time, check count and observer URL. Agents can read the same records with the list_egress_ip_history MCP tool and optionally filter by IP family.
Lockdown Mode stops the poll before DNS or connection setup and prevents an in-flight result from being written. No public-IP observation is saved while global or project Lockdown Mode is effective. Polling resumes automatically after Lockdown Mode is turned off if the feature remains enabled.
Pinning traffic to the observed public IP
Only communicate from this public IP pins the project to the addresses the observer confirmed in its latest round. After confirmation, proxy, Replay, Fuzzer and raw TCP/UDP/DNS traffic leaves only while the pin is confirmed:
- While pinned, the observer runs every 30 seconds and a confirmation is trusted for 45 seconds. An expired confirmation, an observer failure, a different address, or an address in a family that was not pinned blocks traffic until a matching observation arrives.
- Local interface changes (VPN up/down, Wi-Fi switch) and upstream route changes invalidate the confirmation at once and trigger an immediate re-check. An observation that started on a different network is ignored.
- The guard refuses before DNS lookup or socket creation. When traffic stops being allowed, live proxy connections, held Intercept requests, Replay and Fuzzer runs are closed. A blocked attempt shows an error toast.
The pin is stored in the project database, so exports carry it. It cannot be turned on while Lockdown Mode is active, and a pinned project blocks all user traffic while Lockdown Mode is active because the observer cannot run. The pin can always be removed. Public IP recording cannot be disabled while pinned. The decision reflects what the observer sees; a target reached through a different route (for example a split tunnel) may see another address.
Looking like a browser
Lanius terminates TLS, so the server fingerprints the proxy rather than your client. Settings > TLS fingerprint reshapes the handshake to resemble Chrome, Firefox or Safari, or forces TLS 1.2. The cipher manager lists the suites supported by the engine and lets you add, remove, search and reorder them without writing an OpenSSL expression. Recommended suites are shown by default; expand the full list only when testing legacy compatibility. This covers the cipher list and TLS version, not a full JA3 or JA4 match.
On Windows, macOS and Linux, HTTPS server certificates use the operating system's trusted roots automatically, including trusted VPN and private CAs. A CA already trusted for TLS server authentication does not need to be registered again in Lanius. Restart Lanius after changing the Windows or Linux certificate store; new macOS connections use current Keychain decisions. Hostname and leaf validity checks remain enabled, and verification does not download intermediates or revocation responses.
Settings > Proxy > Upstream CA certificates lets a project add other VPN or private CAs when needed. Choose a PEM file or paste its public CA certificates, then apply. These supplement system trust on Windows and macOS; other platforms extend the default public CA list. Server certificate verification stays enabled by default. Applying or removing it restarts proxy connections and cancels active Replay/Fuzzer requests. CA certificates are saved in the project database, but importing a project keeps the current CA trust: new trust anchors must be registered explicitly. Private keys, leaf certificates and expired CAs are rejected.
Settings > Proxy > Upstream certificate exceptions is for a host whose certificate is known to be wrong, such as a hostname mismatch, that you are authorized to test. A trusted CA cannot fix a mismatch. List the exact hostname or IP address (no wildcards), or use Allow this host on the failed request in HTTP history, and Lanius accepts any certificate from that host in this project; every other host stays verified. When verification fails, the browser gets a Lanius page that names the certificate error instead of a bare 502, and the request detail shows the same error. Exceptions are saved in the project database, but importing a project never adds them.
For authorized labs with many deliberately invalid certificates, the same section can disable verification for every upstream HTTPS server after a warning confirmation. This accepts all certificate failures, including unknown issuers, hostname mismatches and expired certificates, not only self-signed certificates. The choice is saved for the current project, but importing a project never turns it on or off.
Media body storage
Settings > Project > Media body storage skips storing HTTP image, audio and video bodies at or above a configurable size. The default is 5 MB (5 × 1,024 × 1,024 bytes); set 0 for unlimited storage. The limit applies to the actual captured body bytes, including uploads and modified request snapshots. Content-Type identifies media; missing or generic binary types also use common media filename extensions. HTML, JSON, other non-media downloads and raw TCP streams retain their bodies.
Headers, status, timing and original sizes remain in History, with a notice when the body was omitted. Omitted bodies are excluded from body search and previews. Proxy forwarding remains complete. The setting is saved with the project and takes effect on new captures immediately. Existing captures and imported backups are not trimmed automatically.
System capture
Some applications never look at the system proxy. Lanius can capture them at the operating system level instead, so they do not need to be configured at all. This is experimental.
Open the Settings tab, find System capture, and choose whether to capture every application or only the ones you name. Selected applications gives you a list: add a rule and type part of a name, or pick from what is running and have its path filled in. Each rule matches part of the executable path, so a fragment is enough:
chrome Google Chrome, and anything with chrome in its path
/Applications/ everything installed there
pid:4123 one processSet a rule to exclude to leave something out, or untick it to switch it off without deleting it.
On macOS this installs a network extension the first time, and macOS will ask you to approve it in System Settings > General > Login Items & Extensions > Network Extensions. Lanius shows an approval warning with a button that opens those settings directly. On Windows the helper needs to run with administrator rights. On Windows and Linux, Lanius checks the native redirector requirements before saving the setting. If elevation, kernel support, or another platform requirement is missing, Settings shows the exact reason instead of macOS approval instructions.
The system redirector cannot be reconfigured while it is running, so changing the setting after capture has already started takes effect on the next launch. Lanius says so when that happens.
Applications that pin their certificates will refuse the connection rather than be intercepted. That is a property of pinning, not something Lanius can work around.
Appearance and language
Settings > Appearance offers Steel as the default dark palette, the original dark palette as Classic Dark, Light, Copper, Sage, Paper, and six editor palette adaptations: Monokai Classic, Gruvbox Dark, Dracula, Tokyo Night, Catppuccin Mocha, and Solarized Dark. Their sources are credited in theme-sources.md. It follows the system by default.
The interface font and the editor font are set separately, with their own sizes, since one is for labels and the other for raw HTTP where a monospace font keeps columns lined up. A preview shows the result as you change it.
The interface is available in English and Korean. Switch in Settings; the choice is remembered. Settings and Docs use the available window width when you resize the app.
Keyboard shortcuts
Switch between Lanius screens with ⌘+Option+0–9 on macOS or Ctrl+Alt+0–9 on Linux and Windows (Dashboard is 0, Proxy is 1, and the remaining screens follow the tab bar). Use ⌘+Option+D or Ctrl+Alt+D for Docs. Send the selected request to Replay with ⌘+R on macOS or Ctrl+R on Linux and Windows, or to Fuzzer with ⌘+I / Ctrl+I. These work in Proxy history, Intercept and Target. In Replay, Send to Fuzzer carries the current edited request; in Fuzzer, Send to Replay opens the selected result. An open context menu uses its right-clicked request. While the Intercept view is open, toggle interception with ⌘+Option+I on macOS or Ctrl+Alt+I on Linux and Windows, and forward the selected message with ⌘+Shift+Enter or Ctrl+Shift+Enter. Drop is also available as a configurable action, but has no default binding to prevent accidental use. Settings > Shortcuts lists these actions alongside every screen and Replay action, and lets you record, disable, or reset each shortcut. These choices are saved on this machine.
Updates
Settings, About says which commit a build came from and checks whether a newer one has been published. Nightlies are compared by commit and stable releases by version. The desktop app can install what it finds, in one press; nothing downloads on its own, and the check itself can be turned off for a network where nothing should leave the machine.
Signed builds can install themselves: Install and restart downloads the new bundle from the selected stable or nightly channel, verifies its signature, replaces the app and comes back up. Windows MSI and NSIS installs each receive their matching installer. Linux AppImage installs support this; .deb installs use the download link for upgrades. The README describes the signing setup.
Lockdown Mode
For sensitive networks, Settings > Lockdown Mode has separate global and project switches. They block Lanius-owned external requests while browser and proxy traffic remain available by default. A project option can additionally drop out-of-scope HTTP/HTTPS proxy traffic and all raw TCP/UDP while Lockdown is active. The global switch can also be forced before launch with LANIUS_LOCKDOWN=1. See lockdown-mode.md for the exact scope.
