Client-side Javascript Functions

The client API mirrors RAGE:MP's Client-side functions page. It has three layers, and knowing which layer a call comes from tells you what to expect from it.

Layer	What it is	Where it lives
mp.game.*	Every GTA native, wrapped and typed	mp_natives.js, generated
Class methods	Player / Vehicle / Entity / Object / Blip / Camera methods	mp_classes.js, generated
Platform API	Pools, events, GUI, voice, storage, Vector3, labels, markers	the client itself

Current coverage: 5424 of 5733 members — 1294 class/namespace members plus the 4439 mp.game.* functions. The live breakdown, including the exact list of what is still missing, is regenerated by python tools/audit-class-api.py into docs/RAGEMP_CLIENT_API_STATUS.md, and the gap list with reasons into docs/NOT_IMPLEMENTED.txt.

Both the documented name and the native's full name work, so either of these is fine:

mp.game.ped.addArmourTo(ped, 50);      // the name the wiki documents
mp.game.ped.addArmourToPed(ped, 50);   // the native's own name

---

mp.game.* — the natives

Every GTA native is available, camel-cased under its namespace, with a typed signature:

mp.game.ped.setPedArmour(mp.players.local.handle, 100);
mp.game.graphics.drawMarker(1, x, y, z, 0,0,0, 0,0,0, 1,1,1, 255,0,0,128, false, false, 2, false, '', '', false);

These take the entity handle, not the entity object — that is the native's own signature.

Class methods

RAGE:MP's class methods are the same natives with the handle supplied for you:

// identical calls
mp.game.ped.addArmourToPed(mp.players.local.handle, 50);
mp.players.local.addArmourTo(50);

They are generated by mapping each name the wiki lists onto the native that implements it (ADD_ARMOUR_TO_PED → addArmourToPed → Player.addArmourTo). A method is only generated when that lookup resolves and the native's first parameter can actually take a handle, so nothing is bound on a coincidental name match.

Hand-written methods always win: where the client has a richer implementation (deferred spawns, server-id resolution, freeze-state tracking), the generated forward is skipped rather than clobbering it.

Method availability

If a method is missing, it is missing for one of three reasons:

1. RAGE:MP uses different wording for the same native. Its API was named against an older native list, so Vehicle.setNeonLightEnabled is SET_VEHICLE_NEON_ENABLED (no "LIGHT"). Most of these are now resolved automatically through a name→hash lookup, with anything left over going one verified line at a time into tools/ragemp-name-aliases.json.
2. RAGE:MP implements it itself, not through a native. The Player.setVoiceFx* family, P2P voice (p2pEnabled / p2pConnected) and the object glow styles are RAGE:MP engine features with no GTA native behind them.
3. The native genuinely isn't in our table. mp_natives.js is built from a named-natives-only dump (6673 natives, no unnamed _0x… entries), so anything the dump omits has nothing to forward to.

---

Vector3

mp.Vector3 is a value type with full maths. Binary operations return a new vector and never mutate the receiver:

const dir = target.position.subtract(mp.players.local.position).unit();
const dist = target.position.dist(mp.players.local.position);

Available: add, subtract, multiply, divide, negative, clone, dot, cross, length, unit, equals, max, min, toArray, toAngles, angleTo.

Every vector the client hands you is a real Vector3 — including entity.position, raycast results and Vector3 event arguments — so you can chain straight off a position read.

---

Entity pools

mp.players, mp.vehicles, mp.peds, mp.objects, mp.pickups, mp.blips, mp.markers, mp.checkpoints, mp.colshapes, mp.labels, mp.browsers, mp.cameras, mp.dummies.

Each supports at, atRemoteId, atHandle, exists, forEach, forEachInStreamRange, apply, toArray and length. Streamed pools resolve by server id; toArray() unions the streamer's live set with locally created entities, so a streamed remote vehicle is visible to forEach even though nothing was pushed into the pool by hand.

Every entity also carries dist(position), distSquared(position), getVariable, setVariable, hasVariable and resetVariable.

---

Labels and markers

Both are draw-every-frame primitives — the client draws them for you once created:

const label = mp.labels.new('Hardware Store', new mp.Vector3(x, y, z), {
    los: false, font: 4, drawDistance: 20, color: [255, 255, 255, 255], dimension: 0
});

const marker = mp.markers.new(1, new mp.Vector3(x, y, z), 1.5, {
    color: [255, 128, 0, 120], visible: true, dimension: 0
});
marker.setColor(0, 255, 0, 200);

Labels respect drawDistance (a squared compare before any native runs, so hundreds of labels only cost you the near ones) and los, which does a real raycast — only for labels already in range.

---

Checkpoints and colshapes

const cp = mp.checkpoints.new(1, new mp.Vector3(x, y, z), 3.0, { color: [255,128,0,120] });
const zone = mp.colshapes.newSphere(x, y, z, 15.0);

Both raise real enter/exit events — see Client-side events for playerEnterCheckpoint and playerEnterColshape. All six colshape variants use real geometry, and a shape only collides with a player in the same dimension.

---

Namespaces

Namespace	Notes
mp.console	logInfo / logWarning / logError / logFatal / clear / reset / verbosity
mp.events	add, remove, call, callRemote, callRemoteProc, addProc, addDataHandler; callRemoteUnreliable is an alias — our transport is reliable-only
mp.gui	chat.push / chat.activate / chat.show / chat.safeMode / chat.colors, cursor.show, execute
mp.keys	bind, unbind, isDown, isUp
mp.raycasting	testPointToPoint, testCapsule, testPointToPointAsync (resolves immediately — our shape test is synchronous)
mp.storage	data (persists across reconnects), sessionData, flush()
mp.voiceChat	muted, defaultVolume, lastVad, minVad, preprocessing params
mp.system	isFocused, isFullscreen, notify()
mp.user	preferences.language, preferences.lowQualityAssets
mp.nametags	enabled, orderByDistance, useScreen2dCoords, set(), update()
mp.discord	update(); requestOAuth2() rejects — there is no OAuth2 broker here

---

How a RAGE:MP name finds its native

Most methods resolve by stripping the class token (ADD_ARMOUR_TO_PED → Player.addArmourTo). The rest are names RAGE:MP inherited from an older community native list, where nothing in the string connects the two. Those are resolved by hash:

RAGE:MP name  →  hash (community native DB)  →  our generated wrapper

This target is RAGE:MP, not FiveM. The native database is used as a name→hash dictionary and nothing more — the API surface comes exclusively from the RAGE:MP wiki's own member list (tools/ragemp-class-api.json), so a native that RAGE:MP does not document is never bound, and argument shapes are verified against RAGE:MP's own per-member pages. The two projects happen to share native naming; that overlap is the only thing used.

Three guards apply: the name must match exactly one native (ambiguous names are skipped), the hash must already exist in our table, and for a class method the native's first parameter must be that class's own entity type. That last one matters — without it DESTROY_CAM reduces to destroy and would bind Entity.destroy.

Signature fidelity

Existence is not correctness. Each generated forward assumes the RAGE:MP signature is the native's minus the handle, and that is checked against every member's own wiki page by python tools/verify-class-signatures.py:

917 signatures checked, 859 match, 58 do not. The full list is in docs/CLASS_SIGNATURE_MISMATCHES.md. Two kinds:

• Out-parameter getters (25) — the native returns values through pointers. Our wrapper unpacks them into an array; RAGE:MP returns a named object. So veh.getColours()[0] works but veh.getColours().colorPrimary is undefined.
• Trailing defaults (33) — the native takes extra trailing arguments that RAGE:MP defaults. Calling with the documented argument count leaves them undefined, which the bridge pushes as 0/false — usually the right default, but not guaranteed per call.

Regenerating

After changing the client's class surface, or after refreshing mp_natives.js:

python tools/gen-class-api.py     # rebuild mp_classes.js
python tools/audit-class-api.py         # refresh docs/RAGEMP_CLIENT_API_STATUS.md
python tools/verify-class-signatures.py # re-check signatures against the wiki

---

mp.voice — the alt:V Voice API

A second facade over the same 3D voice engine, using alt:V's member names. RAGE:MP's own voice surface (Player.setVoiceFx*, P2P voice) is deliberately not implemented — the voice stack here is alt:V-derived, so this is the shape it exposes. mp.voiceChat (RAGE:MP shape) keeps working and both drive the same engine, so they can be mixed.

mp.voice.toggleInput(true);              // start transmitting
mp.voice.muteInput = false;
mp.voice.noiseSuppressionEnabled = true;

// voice activation instead of push-to-talk
mp.voice.activityInputEnabled = true;
mp.voice.activationLevel = 0.08;         // 0..1, compared against the measured mic level

for (const d of mp.voice.getAvailableInputDevices())
    mp.console.logInfo(`${d.uid}: ${d.name}`);
mp.voice.inputDevice = 'Microphone (USB Audio)';

Member	Notes
activationKey	read-only; the key a UI treats as push-to-talk
activationLevel	0..1 threshold for voice activation
inputDevice	device name, or null when there is no capture device
muteInput	inverse of "transmitting"
noiseSuppressionEnabled	RNNoise on/off
voiceControlsEnabled	read-only; true once the voice engine came up
activityInputEnabled	voice activation on/off
getAvailableInputDevices()	[{ uid, name }]
toggleInput(enabled)

activityInputEnabled is real voice activation, not a stored flag: enabling it holds the capture device open (you cannot measure a level on a closed mic) and opens the transmit gate whenever the mic level crosses activationLevel, with a 300 ms hangover so the gate does not slam shut between syllables.

alt:V gates several of these behind an "Extended Voice API permission" prompt. There is no permission broker here, so they are simply available.