ZeroDev Sentinel docs
ZeroDev Sentinel: install and how to use
ZeroDev Sentinel is a server-side anti-cheat resource for FiveM. It records what the server sees, judges it with rules that explain themselves in plain sentences, and gives your staff the evidence before anyone is punished. It starts on the protection level Watch: it logs and punishes nobody until you choose otherwise. This guide covers version 0.1.2 on ESX, QBCore, Qbox and standalone servers.
1. Requirements
- OneSync on (
set onesync onin server.cfg). The resource declares OneSync as a dependency, so FXServer does not start it without OneSync. - An FXServer artifact that runs Node 22 scripts (any current recommended artifact; the manifest asks for
node_version '22'). Node does the hashing, compression, file listing and the appeal web route. - Optional: oxmysql. With it, Sentinel keeps its records in your database and creates its own tables on first start. There is no SQL file to import. Without it, everything lives in the resource's own key-value store.
- Optional: pma-voice, for listening to a player's voice chat from the console or the browser panel.
The framework matters for two things only: the money and item trail (money and items that no shop, job or script paid out, reward loops, rollbacks) and who your staff are (ACE groups, which every framework uses). Detection, cases, clips, the console, the browser panel, Discord and the owner desk work the same way on every framework.
2. Five-minute install
Nothing needs SQL, no config edit is needed for the first start, and no other resource is changed.
Step 5: Open the owner desk
Press F7 in game (or type sentinel:console in F8) and open OWNER. The First run sheet lists what is done and what is not: Discord, clip storage, your staff and their permissions. Each line has one button. Then play for a few days on Watch, read what Sentinel found on the Right now sheet, and switch the protection level to Recommended on the same desk (section 6).
3. Your framework
Sentinel finds the framework and the inventory on its own when it starts. There is nothing to set. sentinel:trail in the server console shows which adapter it chose.
ESX Legacy
set onesync on
ensure oxmysql
ensure es_extended
# your other resources: inventory, jobs, shops ...
ensure zerodev_sentinel- Money and items: read from ESX's own server events (account money added, removed or set; inventory items added or removed; items given from player to player, with both sides linked). Rollback uses ESX's own player methods and reads the result back.
- ESX with ox_inventory: money is followed the same way. In this mode ESX's item methods do not report results, so Sentinel decides only from the read-back.
- Staff: give your admins the ACE permission (section 5). ESX's own
group.adminworks as an ACE group when you add the line for it.
QBCore
set onesync on
ensure oxmysql
ensure qb-core
ensure qb-inventory # or ox_inventory
# your other resources ...
ensure zerodev_sentinel- Money: read from QBCore's money-change event (add, remove and set, with the reason). A loss and an equal gain in the same moment are linked as a transfer, because QBCore has no transfer event of its own.
- Items: with ox_inventory, items are followed through its hooks. With qb-inventory, item changes are not followed (its listener payload is not documented), but a rollback can still move items with qb-inventory's documented exports.
- Staff: QBCore's permission groups are ACE groups (
group.admin,group.god). Add the ACE lines in section 5.
Qbox
set onesync on
ensure oxmysql
ensure ox_lib
ensure qbx_core
ensure ox_inventory
# your other resources ...
ensure zerodev_sentinel- Money: read from Qbox's money-change event; balances are read back with
qbx_core's own exports. Money changed for an offline character (by citizen id) fires no event, so it is not in the trail. - Items: Qbox uses ox_inventory, and Sentinel follows its items through ox_inventory's hooks. Items that stand for framework money are left to the framework part.
- Staff: ACE groups, as on QBCore.
Standalone (no framework)
set onesync on
ensure zerodev_sentinel- Everything except the money and item trail works: movement, world, events, bait events, godmode, state bags, weapons, cases, clips, replay, console, panel, Discord, roles and bans.
- With ox_inventory and no framework, items are followed through ox_inventory.
- Storage: the key-value store, or oxmysql if you start it first.
Checking it works
sentinel:health: the startup box again. Every row is a real check; READY means nothing is missing.sentinel:diag: why storage, adapters and the level were chosen, every rule with its counters, and the latest decisions.sentinel:trail: the money and item adapter in use, and the last changes of one player.
In auto storage mode Sentinel uses oxmysql only when oxmysql is already started (it waits up to 15 seconds while oxmysql is starting); if oxmysql starts later, it uses the key-value store.
4. First run: the owner seat
The first start prints a one-time code in the server console, valid for 30 minutes (sentinel:owner code prints a new one). Join and type /sentinel owner <code>: the seat is bound to your account and the code dies.
Only the owner seat can change keys, the protection level, roles and permissions. sentinel:owner reset in the server console frees the seat if you have to move it to another account.
5. Staff and permissions
On the owner desk, STAFF: pick an online player, pick a role, press GRANT. Four roles ship:
| Role | Can, out of the box |
|---|---|
| Support | see cases, handle reports, the duty map |
| Moderator | as Support, plus replays, kick and quarantine |
| Admin | as Moderator, plus reveal, clear, live view, ban requests, appeals |
| Senior admin | as Admin, plus a ban that applies at once |
Permissions are yours to choose. Every role is a list of switches you can change. Make your own roles from a copy (for example "Trial mod"), rename any role, and give one person an extra switch or take one away without changing their role. The switches marked dangerous (ban at once, listen to voice chat, settings, staff management) ask for a second press. Every change is written to the staff record, so you can always see who changed what.
ACE still works for servers that already use it, and is read on every check:
add_ace group.admin sentinel.admin allow
add_ace group.moderator sentinel.staff allow
add_principal identifier.fivem:1234567 group.admin| Permission | Role it gives |
|---|---|
sentinel.admin |
Admin (Senior admin until the owner seat is claimed) and every sentinel:* command in game |
sentinel.staff |
Moderator; movement rules treat a staff member's own movement as explained |
A few commands (sentinel:audit, sentinel:case, sentinel:replay, sentinel:save) are registered as restricted: in game they also need command.<name>. The common line add_ace group.admin command allow covers that. The server console can run every command.
6. The protection level
One switch on the owner desk (OWNER, PROTECTION), also sentinel:level in the server console and Config.Protection in config/config.lua. What you set on the desk is saved and wins over the config line after a restart.
- Watch: every rule watches and writes what it would have done; nobody is alerted and nothing happens to anyone. Meanwhile Sentinel learns what is normal on your server: the event observer (rates and shapes of each event) and the join check learn only on Watch; entity, combat and money limits keep learning from clean play.
- Recommended: every rule is live. Staff get alerts with the evidence, and a case opens. Sentinel never punishes on its own; your staff decide.
- Strict: as Recommended, and a high score quarantines the player until staff look; a proof-level finding (Tier A: a bait event fired, a forged request) also queues a ban for the ban wave.
Under the switch the desk lists every rule with its state and where that state comes from. Pin a single rule to Live or Shadow there if your server needs it (a rule that keeps flagging one of your own scripts, for example); a state line in Config.Rules pins it in the config file and wins over the desk.
A sensible order:
- Watch for a few days of normal play.
- Read what the rules would have done: the Right now sheet,
sentinel:diag rules(the shadow column counts the times a rule would have scored) andsentinel:diag decisions 100. - Pin to Shadow any rule that flagged someone you trust, and tell us about it.
- Switch to Recommended. Staff now see cases and evidence; nobody is punished automatically.
- Strict when you trust what Recommended shows you.
To step back at any time: Watch on the desk, or sentinel:level watch.
7. The staff console (F7)
Press F7 in game (or type sentinel:console in F8). Each player can change the key in the game's settings (Key Bindings, FiveM). Esc closes it.
- Right now says in plain words what is happening on your server: who looks like a threat and why, what was proven, what was explained (a revive, a script's own teleport) and what needs you. Each player who matters gets one row: what they did, what it probably is, how sure Sentinel is (Unusual, Suspicious, Very likely, Proven), what Sentinel already did, and one next step.
- Case files: open cases first, each with its finding, the evidence, screen clips, the seal (whether the case's hash chain still verifies) and the keys REVEAL, QUARANTINE, CLEAR and BAN (CLEAR and BAN ask for a second press).
- Duty map: online players on a 3D model of the island. The terrain is measured by the first staff member's game and kept in
data/map/. - Play replay closes the console and plays the incident back in a private world; press Backspace to return.
- Owner (owner seat only): First run, Protection, Keys, Staff, Internal affairs and Server check.
Commands typed in game go in F8 without a slash, or in chat with a slash (/sentinel:diag).
8. The browser panel
The same desk in a web browser, for when you are not in game:
- on the server PC:
http://127.0.0.1:<port>/zerodev_sentinel/panel/(the port is your server port, 30120 by default) - from elsewhere:
http://<server address>:<port>/zerodev_sentinel/panel/, or your own HTTPS address in front of it
Signing in. Use a one-time code: type /sentinel panel in game (the code is shown to you only, works once, for 5 minutes), or run sentinel:panel <id or account> in the server console. A session lasts 8 hours; SIGN OUT and END ALL MY SESSIONS end it. Your role and switches apply in the browser exactly as in game.
Live view. Staff with the live view switch can watch any online player's screen from the panel or the console. Nothing is kept on disk unless you press Save to case.
Listening. Listening to that player's voice chat needs the listen switch (owner only by default) and pma-voice. The sound plays through your own game, so be in game on the same server.
Every live view and every listen is written to the staff record and to the player's case, and players are told on their first join that staff may watch and listen (the recording notice, Config.Notice).
9. Cases, quarantine, kick, ban and appeals
When something is found, Right now names the player in plain English: what they did, how sure Sentinel is, and what it did about it. Open the case to see the evidence (positions, screen clips sealed with SHA-256, a 3D replay) and use the keys:
- REVEAL: show the player's identifiers.
- QUARANTINE: move the player out of the real world into a private routing bucket until staff look.
sentinel:release <server id>brings them back. - CLEAR: no cheat; closes the case and stops its queued ban.
- KICK: off the server, with a reason the player reads.
- BAN: through the ban wave, with an appeal code.
The ban wave. Bans, including a BAN pressed in the console, are queued for the ban wave and applied only on Strict, at a random time 2 to 24 hours later by default (Config.Actions). sentinel:wave lists them, sentinel:wave now <account> applies one at once, sentinel:wave cancel <account> cancels it.
Appeals. A banned player sees an appeal code in the refusal message (8 letters and digits). They cannot join to appeal, so the appeal reaches you from outside the game:
- Through your staff. The player sends the code to your staff (a Discord ticket, mail). A staff member records it with
sentinel:appeal <code> <text>. - Through the web route. Sentinel answers on your server's own address:
http://<server address>:<port>/zerodev_sentinel/appeal. It accepts only a POST with JSON{ "code": "...", "text": "...", "contact": "optional" }(text 10 to 2000 characters), so it is meant for a form on your website, not a page to send players to. Every accepted request gets the same answer, 202 "received", whether the code matched a ban or not. Too many requests from one address (5 in 10 minutes) or in total (30 a minute) get 429. Behind your own HTTPS reverse proxy, setset sentinel_http_proxy trueso the limits count the visitor's address.
The refusal text does not say where to appeal. Tell players in your rules, or write your own text with your link in Config.Bans.message (keep the %s, it becomes the code), for example 'You are banned. Appeal at discord.gg/yourserver with code %s'.
Staff decide in the console or with sentinel:decide <case or code> accept|reject [note]. Accept lifts every active ban of that account or case and cancels a queued wave ban; reject keeps the ban. A decision is final and is written to the case log. sentinel:appeals lists the open ones.
10. Reports and internal affairs
Player reports. Any player types /report <server id> <what happened> to report a player or a staff member. The last two minutes around both players are saved into a case. Staff close a report as actioned, no breach, no proof or duplicate. Reports about staff go to the owner only.
Internal affairs (owner only, on the owner desk): what each staff member did, and what needs a look (grants to themselves, hiding in combat, bans with no case, teleports to players). A weekly fairness report is written to data/reports/; sentinel:staff report writes it now. Nothing in the staff record punishes staff.
11. Guarding a resource
Without any setup, Sentinel observes every event players can send to your scripts and judges the player. It cannot stop another resource's handler from running. To stop a bad call before a script runs, add the guard to that script.
sentinel:auditlists the events of every started resource that clients can trigger, graded by what their handlers do (for example a handler that gives money with an amount the client picks). Use it to decide which resources to guard first.sentinel:guard <resource>prints the line to add and the resource's server_scripts as they would read. The guard line must be the first server script. The command only prints; it changes no file.
lua
server_scripts { '@zerodev_sentinel/guard/guard.lua', 'server/main.lua' }
- Restart that resource.
sentinel:guard <resource>now says "guarded since ...".sentinel:guardalone lists every guarded resource. - Add rules with
sentinel:patch. Until a resource has rules, its guard lets every call through, except calls from quarantined players, which every guard drops. Example: at most 3 buys in 10 seconds, an item name of up to 32 characters and an amount from 1 to 10; a refused call counts as proof ("hard": true):
text
sentinel:patch myshop myshop:buy {"args":[{"type":"string","maxLen":32},{"type":"number","int":true,"min":1,"max":10}],"rate":3,"window":10000,"hard":true}
For a framework callback, write callback:<name> as the event: sentinel:patch es_extended callback:esx:getPlayerNames {"args":[{"type":"table","maxItems":4}]}. sentinel:patch <resource> lists a resource's rules; sentinel:patch <resource> <event> off removes one.
The guard never edits files and never changes what a handler receives. It is plain, readable Lua in guard/guard.lua.
12. Discord and clip storage keys
The easiest way is the owner desk, KEYS: paste each value and press SAVE. Keys are kept on the server only; the desk and every reply say only "set" and never show a key again.
| Line on the desk | What it is | Or in server.cfg (use set, never setr) |
|---|---|---|
| Discord bot token | a bot you made at discord.com/developers, invited to your server | set sentinel_discord_token "..." |
| Discord server id | right-click your server, Copy Server ID (developer mode on) | set sentinel_discord_guild "..." |
| Alert channel id | where cases are posted, with buttons for your staff | set sentinel_discord_channel "..." |
| Appeals channel id | where appeals are posted | set sentinel_discord_appeals "..." |
| FiveManage key | a media key from fivemanage.com; clips are uploaded there | set sentinel_fivemanage_key "..." |
| Clip webhook | instead of FiveManage: a Discord webhook that receives the clips | set sentinel_clip_webhook "..." |
| Clip storage | fivemanage, discord, off, or automatic (the first one that is set) | set sentinel_clip_storage "..." |
A key in server.cfg wins over the desk, and the desk says so. Discord buttons follow the same roles and switches as the console. Without any of these, cases and clips stay on your server (data/) and the console and panel show them.
The simpler webhook alerts still work: set sentinel_webhook "https://discord.com/api/webhooks/<id>/<token>" or sentinel:webhook set <url> in the server console. sentinel:webhook test sends a test message; sentinel:webhook shows the state without ever printing the URL; sentinel:webhook clear removes the stored one.
13. What each rule catches
Every rule writes a plain sentence when it fires, saying what it saw and what it ruled out. Before a rule counts against anyone, Sentinel checks the reasons that would explain it (a respawn, a revive, a staff action, one of your own scripts that moves or pays players). An explained finding is logged and never scored.
- Tier A is proof: something only a cheat does (a bait event that no real script calls, a forged request).
- Tier B is strong: a staff member should look. On its own it opens a case only when two different families agree, or the same one repeats.
- Tier C is weak: it adds to the score and never acts alone.
Movement
| Rule | Catches | Explained by | Tier |
|---|---|---|---|
| teleport | A jump of over 150 m (or 3x what is possible) between two samples | spawn, respawn or revive, routing bucket change, staff, your script's teleport event | B |
| foot_speed | Running over 12 m/s for 1.5 s on foot, not falling, not ragdoll | spawn, respawn, staff, your script | B |
| air_float | Moving while frozen (noclip); "falling" without coming down (fly); moving fast by coordinates while the game says the player stands still (speed menus) | spawn, respawn, staff, your script | B |
| foot_rise | Climbing straight up faster than 4 m/s on foot for 3 s | spawn, respawn, staff, your script | B |
| vehicle_speed | A vehicle over 1.3x the fastest its class reaches on your server | staff, your script | B |
World
| Rule | Catches | Explained by | Tier |
|---|---|---|---|
| entity_blocked | Spawning a model you blocked | your scripts that spawn it | B |
| entity_rate / entity_model_rate | Spawning far more entities (or one model) per minute than players and scripts do | your spawning scripts | B |
| entity_untrusted | Entities created by a client with no script that spawns them | your spawning scripts | B |
| entity_disguised | A burst of the same ambient ped model around one player (a menu disguising its spawns as traffic) | spawn, respawn, bucket change | B |
| explosion_weapon / explosion_rate / explosion_hidden / world_distance | Explosions the held weapon cannot make, too many, invisible ones, or far away from the player | your scripts' explosions | B |
| projectile_weapon / projectile_rate | Projectiles the weapon cannot fire, or too many | B | |
| ptfx_rate | Particle effect floods (a common crash and lag attack) | B | |
| fire_rate | Fires started faster than any weapon allows | B |
Combat
| Rule | Catches | Tier |
|---|---|---|
| damage_plausibility | Damage over the weapon's possible value | B |
| hit_range | Hits from beyond the weapon's range | B |
| shot_rate / kill_rate | Firing faster than the weapon; kills per minute far over your server | B |
| silent_aim | Hits where the player was not aiming, repeated | B |
| aim_stats / aim_stats_backed | Headshot share and snap angles far outside your server's players over 30+ fights | C (B with another family) |
| esp_los_backed | Repeated aiming at players the witnesses report hidden behind walls | C to B |
Events and requests
| Rule | Catches | Tier |
|---|---|---|
| bait_event | Calling a planted event that no real script uses (cheat menus scan for and fire these) | A |
| bait_statebag / statebag_write | Writing state bags only the server should write (making yourself admin, godmode flags), or oversized ones | A / B |
| events_before_ready / _trap | Events sent before the player has even spawned (injected before joining) | B / A |
| event_rate / event_shape | An event called far faster than normal, or with arguments shaped unlike your server's normal calls | B |
| event_flood | 500+ calls in 10 s; on FiveM's own overflow kick, the top events are listed | B |
| guard_rule / guard_hard | Breaking a rule you set on a guarded event (amount, rate, distance) | B / A |
Economy
| Rule | Catches | Tier |
|---|---|---|
| money_nowhere / item_nowhere | Money or items that no shop, job or script handed out (dev events, leaked shop sells) | B |
| economy_repeat | A reward or job event paid 10 times in 30 s (farming by calling it in a loop) | B |
| money_jump / money_rate | A sudden balance jump or a pay rate far over your server | B |
| economy_trap | Claiming a planted reward that no real job pays | A |
| weapon_not_owned | Holding a weapon the inventory never gave | B |
The player's game (witness)
| Rule | Catches | Tier |
|---|---|---|
| godmode | The server reads the player invincible for 30 s (not a shared safe zone) | B |
| witness_backed | Signs from the game backed by the server's own view (godmode, invisibility, no collision) | B |
| witness_bait | The game answering a check it should not know about | A |
| witness_tamper | Sentinel's own checks in the game were hooked or changed | B |
Identity and files
| Rule | Catches | Tier |
|---|---|---|
| ban_fingerprint | A banned player back on a new account (hardware tokens, identifiers, browser fingerprint) | A / flag |
| young_account | A brand-new account (adds weight only) | C |
| scanner_error / too_large | Resources the script scanner could not read, or oversized files worth a look | info |
What no rule promises
Sentinel judges what the server can see and what a player's game reports. It does not read other programs on a player's PC, and no anti-cheat can promise that every cheat is caught. Every finding comes with its evidence, so your staff decide on facts.
14. Commands and configuration
Every command is on the commands page and every config key on the configuration page, both generated from the resource itself. The ones you use most:
| Command | What it does |
|---|---|
/sentinel owner <code> |
claim the owner seat in game with the code from the console |
/sentinel panel |
a one-time sign-in code for the browser panel (staff only) |
/report <server id> <what happened> |
any player reports a player or a staff member |
sentinel:health |
the startup box again |
sentinel:diag |
what Sentinel decided and why |
sentinel:level watch\|recommended\|strict |
change the protection level and save it |
sentinel:wave |
the bans queued for the ban wave |
sentinel:appeals |
the open appeals |
sentinel:audit |
the risky events of every started resource |
sentinel:guard <resource> |
the line that guards a resource |
sentinel:advisor |
the FiveM settings worth changing on your server |
The config files in config/ (config.lua, protection.lua, watch.lua) and the translations in config/locales/ are open to you, as are the framework adapters in bridge/ and the guard file in guard/. Keep them valid Lua and restart zerodev_sentinel after a change. No config edit is needed for the first start.
15. The startup box and the hardening advisor
A moment after storage is ready, Sentinel prints a box in the server console. A FAIL row means Sentinel cannot work properly; a WARN row means it works with less (for example evidence without digests). Print it again at any time with sentinel:health.
| Row | What it checks | When it is not OK |
|---|---|---|
| Build | The version and build label from fxmanifest.lua. | - |
| OneSync | The onesync convar. Sentinel reads positions, entities and game events through OneSync. |
set onesync on in server.cfg |
| Copies running | How many copies of the resource are started, whatever their folder is called. Must be 1. | keep one copy of Sentinel; stop the other folder |
| Storage | Which storage was chosen (mysql or kvp) and whether it is ready. | see sentinel:diag storage |
| Evidence folder | A test write and read in data/archive inside the resource. | create the folder data/archive inside the Sentinel resource and make sure the server can write to it |
| SHA-256 | Whether node/util.js answers; it fingerprints every piece of evidence. | evidence is saved without digests; check the console for a Node error |
| Protection | The protection level, its mode and where it came from. | - |
| Rules | How many rules are live and how many are in shadow. | - |
After the startup box, the advisor prints the FiveM settings worth changing on your server (Config.Advisor.atStartup = false prints it only on sentinel:advisor). For each setting it shows the current value, the recommended one, what it blocks, what it might break, and the line to copy. It reads the values only; it never changes server.cfg. Test a line on your server before you keep it.
16. Languages, and what is recorded
Player and staff texts come from config/locales/<code>.lua, 25 languages in this build. Config.Locale picks the language; with Config.GameLanguage = true a player whose game runs in a language GTA offers reads Sentinel in that language once in game.
Sentinel keeps, for every player, about the last two minutes of positions and events in memory (Config.Recorder), and writes them to an archive in data/archive/ when a case opens or staff use sentinel:save. During a review it asks the suspect's game for screen clips (20-second segments) and stores them in data/clips/. It stores the identifiers FiveM gives the server (an IP address is recorded but never used to refuse a connection by default) and, when Discord runs, the Discord id. On their first join every player gets a short notice in their language that the server records play and that staff may watch and listen (Config.Notice = false turns it off; then tell them in your rules). Mention it in your privacy policy too.
Old archives and clips are not deleted automatically in this build: clear data/archive/ and data/clips/ yourself when you no longer need them.
With oxmysql, Sentinel's records are in the tables zds_account, zds_ban, zds_setting, zds_pair, zds_exemption, zds_case, zds_schema, zds_report and zds_trail. Without oxmysql the same records live in the resource's key-value store (keys starting with zds:). Bans are always mirrored to the key-value store, so they stay enforced while the database is down.
17. Updating
New versions are published to the same Keymaster asset. Sentinel keeps its evidence inside its own folder, so before you replace it:
- Back up the
zerodev_sentinelfolder'sconfig/anddata/folders (archives, clips, reports, learned baselines) and, with oxmysql, thezds_tables. - Stop the server, replace the
zerodev_sentinelfolder with the new one (keep the folder name), and put yourdata/folder back. Move the settings you changed into the new config files rather than copying old files over them. - Start the server and read the startup box (
sentinel:health).
Read the release notes of the new version on the changelog before updating.
18. Uninstall
- Remove the guard line from every resource you guarded (
sentinel:guardlists them) and restart those resources. Do this first: the guard file is gone once the folder is. - Remove
ensure zerodev_sentineland thesentinel.admin/sentinel.stafflines from server.cfg, then stop the resource. Ifsentinel:lockdownchanged entity lockdown, stopping Sentinel puts the original mode back. - Delete the
zerodev_sentinelfolder. That removesdata/with all archives, clips and reports, so copy out anything you want to keep. - With oxmysql, drop the
zds_tables listed in section 16 if you no longer want the bans and cases. The key-value records stay in the server's store under the resource's name and do nothing without it.
Sentinel writes only inside its own folder, its database tables and its key-value store. It does not edit other resources or server.cfg.
Frequently asked questions
Does Sentinel ban or kick anyone on its own?
Not on Watch or Recommended. It starts on Watch, where it only logs what it would have done. On Recommended staff get the evidence and decide. Only on Strict does a high score quarantine a player until staff look, and a proof-level finding queue a ban for the ban wave.
Do I need a database?
No. With oxmysql Sentinel keeps its records in your database and creates its own tables on first start; there is no SQL file to import. Without it, everything lives in the resource's own key-value store.
Does it work on QBCore and Qbox?
Yes. Detection, cases, clips, the console, the browser panel, Discord and the owner desk are the same on every framework. The money and item trail has its own adapter for ESX, QBCore, Qbox and ox_inventory.
How do I give my admins access?
On the owner desk, STAFF: pick an online player, pick a role, press GRANT. Or keep using ACE: sentinel.admin gives Admin and sentinel.staff gives Moderator.
Can a banned player appeal?
Yes. The refusal message shows an appeal code. The player sends it to your staff, who record it with sentinel:appeal, or your website posts it to Sentinel's appeal route. Staff accept or reject it in the console or with sentinel:decide.
Can it see cheat programs on a player's PC?
No. Sentinel judges what the server can see and what a player's game reports. It does not read other programs on a player's PC.
Is listening to voice chat ready?
It needs pma-voice, the listen switch (owner only by default), and plays through your own game.
What do players see?
On their first join, a short notice in their language that the server records play and that staff may watch and listen (Config.Notice). Nothing else changes for a clean player.