This guide will give you a basic introduction to writing a SourceMod plugin. If you are not familiar with the SourcePawn language, it is recommended that you at least briefly read the Introduction to SourcePawn article.

For information on compiling plugins, see Compiling SourceMod Plugins. You can use Crimson Editor, PSPad, UltraEdit, Notepad++, TextPad, Pawn Studio or any other text editor you're comfortable with to write plugins.

Starting from scratch

Open your favorite text editor and create a new empty file. When you have an empty file you can just start writing code using the core language, however, you will not be able to use any of SourceMod features because the compiler does not now about them. This is done deliberately so it is possible to use SourcePawn outside of SourceMod. But since we are writing a SourceMod plugin, it is a good idea to enable access to SourceMod features first. This it done using #include directive. It tells compiler to "paste" the code from another file into yours.

#include <sourcemod>

How does this work? First of all, note that we enclosed file name into angle brackets. Angle brackets tell the compiler to look in the default include directory. By default, it is scripting/include. You can open it right now and see a lot of inc files there. Those are SourceMod include files that describe various functions, tags and other features available for SourceMod plugins. The files are plain-text and you are encouraged to read them. You will notice however, that there's not much code in there, certainly not enough to implement all the great features of SourceMod, so where are they? They are implemented inside a SourceMod core which is written in C++ and is compiled into binary files which end up in bin directory. So how does your SourcePawn code and SM core link together if compiler doesn't know about existence of the latter? SourceMod include files are written specially, so they say that the implementation of functions is somewhere else. Compiler understands that and generate a special code that says that this function call is going outside. When SourceMod loads your plugin, it inspects these bits of code and substitutes it's own internal functions instead. This is called dynamic linking.

Setting up plugin info

Now that we got access to SourceMod features, it is time to setup the information that will be displayed via sm plugins list command. No one likes unnamed plugins. To do that we are going to look inside sourcemod.inc file and see the format that information should be declared. It's always helpful to look inside SM include files to find out information you don't know. There is also an API documentation but it can be outdated and it only has SM core files so if your plugin are going to use any third party extension or another plugin, you will have to study inc files. So, open sourcemod.inc and scroll down a bit until you see this:

/**
 * Plugin public information.
 */
struct Plugin
{
   const String:name[],			/**< Plugin Name */
   const String:description[],	/**< Plugin Description */
   const String:author[],		/**< Plugin Author */
   const String:version[],		/**< Plugin Version */
   const String:url[],			/**< Plugin URL */
};

and this:

/**
 * Declare this as a struct in your plugin to expose its information.
 * Example:
 *
 * public Plugin:myinfo =
 * {
 *    name = "My Plugin",
 *    //etc
 * };
 */
public Plugin:myinfo;

It tells us that we need to create a global public variable myinfo which must be of type Plugin which is a struct with 5 fields which themselves are strings. It may sound complicated for a beginner but it's easy. Let's go ahead and create one:

public Plugin:myinfo =
{
	name = "My First Plugin",
	author = "Me",
	description = "My first plugin ever",
	version = "1.0",
	url = "http://www.sourcemod.net/"
};

The public keyword means that SourceMod will be able to directly access our variable. Plugin: defines a type of our variable. myinfo is, obviously, a name of our variable as required by SourceMod. You see that we initialize it right away. This is preferred way to do when filling out plugin info.

After that the full code of your plugin should look like this:

#include <sourcemod>
 
public Plugin:myinfo =
{
	name = "My First Plugin",
	author = "Me",
	description = "My first plugin ever",
	version = "1.0",
	url = "http://www.sourcemod.net/"
};

Getting code to run

We already include SourceMod features and filled up or plugin info. We now have a perfectly well formed plugin which can be compiled and loaded by SourceMod. However, there is one problem - it does nothing. You might be tempted to just start writing a code after myinfo declaration just to see that it will not compile. SourcePawn, unlike other scripting languages like Lua, does not allow a code to be outside of functions. After reading that, you may probably want to just define some function, name it main probably, compile and load a plugin and see that your code never gets called. So how do we make SourceMod call our code? For this exact reason we have forwards. Forwards are function prototypes declared by one party that can be implemented by another party as a callback. When a first party starts a forward call, all parties that have matching callbacks receive the call. SourceMod declares a plenty of interesting forwards that we can implement. As you can see, forwards are the only way to get our code executed, keep that in mind. So let's implement OnPluginStart forward. As you may have guessed, it is called when our plugin starts. To do that, we'll have to look up the declaration of OnPluginStart. It is declared inside sourcemod.inc, a file we are already familiar with, let's find it:

/**
 * Called when the plugin is fully initialized and all known external references 
 * are resolved. This is only called once in the lifetime of the plugin, and is 
 * paired with OnPluginEnd().
 *
 * If any run-time error is thrown during this callback, the plugin will be marked 
 * as failed.
 *
 * It is not necessary to close any handles or remove hooks in this function.  
 * SourceMod guarantees that plugin shutdown automatically and correctly releases 
 * all resources.
 *
 * @noreturn
 */
forward OnPluginStart();

Empty parentheses tells us that no arguments are passed inside this forward, @noreturn inside documentation tells us that we don't have to return anything, pretty simple forward. So how to write a correct callback for it? Firstly, our callback must have the same name, so it's OnPluginStart, secondly, our callback should have the same number of arguments, none in this case, and lastly, SourceMod needs to be able to call our callback so it needs to be public. So the implementation looks like this:

public OnPluginStart()
{
}

Now we can write code inside curly braces and it will be executed when our plugin starts. Let's output "Hello world!" to server console. To do that we are going to use PrintToServer function. It is declared inside console.inc, however, we don't need to manually include console.inc because it is included automatically as part of sourcemod.inc.

/**
 * Sends a message to the server console.
 *
 * @param format		Formatting rules.
 * @param ...			Variable number of format parameters.
 * @noreturn
 */
native PrintToServer(const String:format[], any:...);

As you can see, this is a native function. It is implemented inside SM core. Judging by it's arguments, we can see that it is a format class function. However, we don't need any formatting right now, so let's just pass "Hello world!" string as an only argument:

public OnPluginStart()
{
	PrintToServer("Hello world!");
}

That's it! The full code of your plugin should look like this:

#include <sourcemod>
 
public Plugin:myinfo =
{
	name = "My First Plugin",
	author = "Me",
	description = "My first plugin ever",
	version = "1.0",
	url = "http://www.sourcemod.net/"
};
 
public OnPluginStart()
{
	PrintToServer("Hello world!");
}

Compile and load your plugin on your server and see for yourself that the message is displayed in server console.

Creating client command

So far we have learned how to set up our plugin and run some code. But our plugin still doesn't do anything useful. Let's add a console command to allow clients to see their kills/deaths ratio. For demonstration purposes, we are going to start with very naive implementation that a beginner could write, then we'll look at the common bugs it contains, fix them and add more advanced features. In the end we'll have a mature feature-rich implementation and experience that will prevent us from making simple mistakes in the future.

Naive implementation

First, we need to register our console command. OnPluginStart is a good place to do that. In order to register our console command, we need to use RegConsoleCmd function, it is declared inside console.inc, here's how it's declared:

/**
 * Creates a console command, or hooks an already existing one.
 *
 * Console commands are case sensitive.  However, if the command already exists in the game, 
 * the a client may enter the command in any case.  SourceMod corrects for this automatically, 
 * and you should only hook the "real" version of the command.
 *
 * @param cmd			Name of the command to hook or create.
 * @param callback		A function to use as a callback for when the command is invoked.
 * @param description	Optional description to use for command creation.
 * @param flags			Optional flags to use for command creation.
 * @noreturn
 * @error				Command name is the same as an existing convar.
 */
native RegConsoleCmd(const String:cmd[], ConCmd:callback, const String:description[]="", flags=0);

This native looks a bit more complicated than the previous one, so let's go through each argument step by step. First one is the name of our command, we'll use "sm_kdr". Second one is a callback, but it has unknown tag ConCmd. In order to correctly call RegConsoleCmd we need to find how ConCmd is declared. Thankfully, we don't need to look hard, it's declared right above RegConsoleCmd:

/**
 * Called when a generic console command is invoked.
 *
 * @param client		Index of the client, or 0 from the server.
 * @param args			Number of arguments that were in the argument string.
 * @return				An Action value.  Not handling the command
 *						means that Source will report it as "not found."
 */
functag public Action:ConCmd(client, args);

This is a function tag. What does it mean? It means that we need to define our own function that will have the same prototype (number of arguments, their tags and return tag) as the one of function tag. However, in this case, we don't need to (and can't) use the same name as in tag, that would result in error. We need to use our own name, let's use Command_KDR. Function tags doesn't require to use the same argument names as specified in declaration, but it's a good idea to use them, because they usually have the most meaning. So, all in all, it leads us to the following definition:

public Action:Command_KDR(client, args)
{
}

Now, if you've been following carefully, you should see that we have another unknown tag here - Action. Let's look how it's declared (core.inc):

/**
 * Specifies what to do after a hook completes.
 */
enum Action
{
	Plugin_Continue = 0,	/**< Continue with the original action */
	Plugin_Changed = 1,		/**< Inputs or outputs have been overridden with new values */
	Plugin_Handled = 3,		/**< Handle the action at the end (don't call it) */
	Plugin_Stop = 4,		/**< Immediately stop the hook chain and handle the original */
};

It's an enumeration tag and in our callback it's up to us to decide which value to return. As you probably understood from reading all the info, SourceMod console commands are implemented as hooks between client and original server code. If we don't return anything and "let it slide" or return Plugin_Continue, the original command from client will reach the server and, since it's highly unlikely that there is sm_kdr command in the game, the message Unknown command "sm_kdr" will be printed to client's console. We don't want that, so let's modify our callback to return Plugin_Handled:

public Action:Command_KDR(client, args)
{
	return Plugin_Handled;
}

It's a very common mistake to forget to return Plugin_Handled so our naive implementation is going to be not so naive after all! Now, after we finally understood how ConCmd tag works, it's time to move to the next argument of RegConsoleCmd. It is a string that will act like a description of our command. However, notice the = after argument name, it indicates that this argument is optional, we can skip it and empty string ("") will be used. And the last argument is flags and it's also optional. Command flags are used to change behavior of command, but for now default behavior is ok. Now we know the meaning of all arguments, let's make a call.

RegConsoleCmd("sm_kdr", Command_KDR, "Displays the client their kills/deaths ratio.");

Notice that we skipped flags because we want default behavior. Now let's see how the full command-related code looks like:

public OnPluginStart()
{
	RegConsoleCmd("sm_kdr", Command_KDR, "Displays the client their kills/deaths ratio.");
}
 
public Action:Command_KDR(client, args)
{
	return Plugin_Handled;
}

Why did we choose sm_kdr as the name of our command? It's handy because it will create convenient chat triggers. The skeleton of our command is ready, it is time to write the actual code for calculating KDR and displaying it to client. Notice that the first argument of our Command_KDR callback is client. When our callback is called, it will hold the index of client who called our command. We'll use this to find necessary information about client and to display that information back. First, let's find the number of kills, we'll use GetClientFrags function for that, it's declared in clients.inc:

/**
 * Returns the client's frag count.
 *
 * @param client		Player's index.
 * @return				Frag count.
 * @error				Invalid client index, client not in game, or no mod support.
 */
native GetClientFrags(client);

It takes one argument - the client index and returns the number of frags. For the only argument of GetClientFrags we'll use client argument of our callback and we also need to store the return value so we'll create a local variable for that:

new kills = GetClientFrags(client);

By this time, it is assumed that you've learned how to read include files and find necessary information, so only links to online SourceMod API will be provided. Next, we need to find the number of client deaths, we'll use GetClientDeaths:

new deaths = GetClientDeaths(client);

Now that we have both kills and deaths, let's find the ratio. Remember that by default all variables in SourcePawn act like integers, but we need the fractional part in this case. Another very important thing to remember is that a result of integer division is also an integer. So we need to convert both kills and deaths to floating point values and store the result in a floating point variable. float function does the conversion.

new Float:ratio = float(kills) / float(deaths);

And now we only need to display the ratio, since our command can be called either via console or chat, it is good idea to display our info accordingly, so we'll use ReplyToCommand. It's another format class function and now we need to apply some formatting. In a nutshell, we pass the string which acts as a template for text and % sign and one-letter type identifier (which is called format specifier) for variable values to put inside that template. After that, we must pass the number of arguments that correspond to the number of format specifiers in our format string. Since we have a floating point number, the format specifier will be %f:

ReplyToCommand(client, "Your KDR is: %f.", ratio);

That's it! Now let's see the full code of our naive implementation:

#include <sourcemod>
 
public Plugin:myinfo =
{
	name = "My First Plugin",
	author = "Me",
	description = "My first plugin ever",
	version = "1.0",
	url = "http://www.sourcemod.net/"
};
 
public OnPluginStart()
{
	RegConsoleCmd("sm_kdr", Command_KDR, "Displays the client their kills/deaths ratio.");
}
 
public Action:Command_KDR(client, args)
{
	new kills = GetClientFrags(client);
	new deaths = GetClientDeaths(client);
	new Float:ratio = float(kills) / float(deaths);
	ReplyToCommand(client, "Your KDR is: %f.", ratio);
	return Plugin_Handled;
}

Compile and test this plugin and notice that something is wrong. We'll talk about it in the next section.

Includes

Pawn requires include files, much like C requires header files. Include files list all of the structures, functions, callbacks, and tags that are available. There are three types of include files:

Core - sourcemod.inc and anything it includes. These are all provided by SourceMod's Core.
Extension - adds a dependency against a certain extension.
Plugin - adds a dependency against a certain plugin.

Include files are loaded using the #include compiler directive.

Commands

Our first example will be writing a simple admin command to slap a player. We'll continue to extend this example with more features until we have a final, complete result.

Declaration

First, let's look at what an admin command requires. Admin commands are registered using the RegAdminCmd function. They require a name, a callback function, and default admin flags.

The callback function is what's invoked every time the command is used. Click here to see its prototype. Example:

public OnPluginStart()
{
	RegAdminCmd("sm_myslap", Command_MySlap, ADMFLAG_SLAY);
}
 
public Action:Command_MySlap(client, args)
{
}

Now we've successfully implemented a command -- though it doesn't do anything yet. In fact, it will say "Unknown command" if you use it! The reason is because of the Action tag. Any command that you type in the console, even if it's registered by SourceMod, will be sent to the engine to be processed. Because we have not had SourceMod block this functionality yet, the engine replies with "Unknown command" because it is not a valid engine command.

public Action:Command_MySlap(client, args)
{
	return Plugin_Handled;
}

Now the command will report no error, but it still won't do anything. This is because returning "Plugin_Handled" in a command callback will prevent the engine from processing the command. The engine will never even see that the command was run. This is what you will want to do if you are registering a completely new command through SourceMod.

Implementation

Let's decide what the command will look like. Let's have it act like the default sm_slap command:

sm_myslap <name|#userid> [damage]

To implement this, we'll need a few steps:

Get the input from the console. For this we use GetCmdArg().
Find a matching player. For this we use FindTarget().
Slap them. For this we use SlapPlayer(), which requires including sdktools, an extension bundled with SourceMod.
Respond to the admin. For this we use ReplyToCommand().

Full example:

#include <sourcemod>
#include <sdktools>
 
public Plugin:myinfo =
{
	name = "My First Plugin",
	author = "Me",
	description = "My first plugin ever",
	version = "1.0.0.0",
	url = "http://www.sourcemod.net/"
}
 
public OnPluginStart()
{
	RegAdminCmd("sm_myslap", Command_MySlap, ADMFLAG_SLAY);
}
 
public Action:Command_MySlap(client, args)
{
	new String:arg1[32], String:arg2[32];
	new damage;
 
	/* Get the first argument */
	GetCmdArg(1, arg1, sizeof(arg1));
 
	/* If there are 2 or more arguments, and the second argument fetch 
	 * is successful, convert it to an integer.
	 */
	if (args >= 2 && GetCmdArg(2, arg2, sizeof(arg2)))
	{
		damage = StringToInt(arg2);
	}
 
	/* Try and find a matching player */
	new target = FindTarget(client, arg1);
	if (target == -1)
	{
		/* FindTarget() automatically replies with the 
		 * failure reason.
		 */
		return Plugin_Handled;
	}
 
	SlapPlayer(target, damage);
 
	new String:name[MAX_NAME_LENGTH];
 
	GetClientName(target, name, sizeof(name));
	ReplyToCommand(client, "[SM] You slapped %s for %d damage!", name, damage);
 
	return Plugin_Handled;
}

For more information on what %s and %d are, see Format Class Functions. Note that you never need to unregister or remove your admin command. When a plugin is unloaded, SourceMod cleans it up for you.

ConVars

ConVars, also known as cvars, are global console variables in the Source engine. They can have integer, float, or string values. ConVar accessing is done through Handles. Since ConVars are global, you do not need to close ConVar Handles (in fact, you cannot).

The handy feature of ConVars is that they are easy for users to configure. They can be placed in any .cfg file, such as server.cfg or sourcemod.cfg. To make this easier, SourceMod has an AutoExecConfig() function. This function will automatically build a default .cfg file containing all of your cvars, annotated with comments, for users. It is highly recommend that you call this if you have customizable ConVars.

Let's extend your example from earlier with a new ConVar. Our ConVar will be sm_myslap_damage and will specify the default damage someone is slapped for if no damage is specified.

new Handle:sm_myslap_damage = INVALID_HANDLE
 
public OnPluginStart()
{
	RegAdminCmd("sm_myslap", Command_MySlap, ADMFLAG_SLAY);
 
	sm_myslap_damage = CreateConVar("sm_myslap_damage", "5", "Default slap damage");
	AutoExecConfig(true, "plugin_myslap");
}
 
public Action:Command_MySlap(client, args)
{
	new String:arg1[32], String:arg2[32];
	new damage = GetConVarInt(sm_myslap_damage);
 
	/* The rest remains unchanged! */

Showing Activity, Logging

Almost all admin commands should log their activity, and some admin commands should show their activity to in-game clients. This can be done via the LogAction() and ShowActivity2() functions. The exact functionality of ShowActivity2() is determined by the sm_show_activity cvar.

For example, let's rewrite the last few lines of our slap command:

	SlapPlayer(target, damage);
 
	new String:name[MAX_NAME_LENGTH];
 
	GetClientName(target, name, sizeof(name));
 
	ShowActivity2(client, "[SM] ", "Slapped %s for %d damage!", name, damage);
	LogAction(client, target, "\"%L\" slapped \"%L\" (damage %d)", client, target, damage);
 
	return Plugin_Handled;
}

Multiple Targets

To fully complete our slap demonstration, let's make it support multiple targets. SourceMod's targeting system is quite advanced, so using it may seem complicated at first.

The function we use is ProcessTargetString(). It takes in input from the console, and returns a list of matching clients. It also returns a noun that will identify either a single client or describe a list of clients. The idea is that each client is then processed, but the activity shown to all players is only processed once. This reduces screen spam.

This method of target processing is used for almost every admin command in SourceMod, and in fact FindTarget() is just a simplified version.

Full, final example:

#include <sourcemod>
#include <sdktools>
 
new Handle:sm_myslap_damage = INVALID_HANDLE
 
public Plugin:myinfo =
{
	name = "My First Plugin",
	author = "Me",
	description = "My first plugin ever",
	version = "1.0.0.0",
	url = "http://www.sourcemod.net/"
}
 
public OnPluginStart()
{
	LoadTranslations("common.phrases");
	RegAdminCmd("sm_myslap", Command_MySlap, ADMFLAG_SLAY);
 
	sm_myslap_damage = CreateConVar("sm_myslap_damage", "5", "Default slap damage");
	AutoExecConfig(true, "plugin_myslap");
}
 
public Action:Command_MySlap(client, args)
{
	new String:arg1[32], String:arg2[32];
	new damage = GetConVarInt(sm_myslap_damage);
 
	/* Get the first argument */
	GetCmdArg(1, arg1, sizeof(arg1));
 
	/* If there are 2 or more arguments, and the second argument fetch 
	 * is successful, convert it to an integer.
	 */
	if (args >= 2 && GetCmdArg(2, arg2, sizeof(arg2)))
	{
		damage = StringToInt(arg2);
	}
 
	/**
	 * target_name - stores the noun identifying the target(s)
	 * target_list - array to store clients
	 * target_count - variable to store number of clients
	 * tn_is_ml - stores whether the noun must be translated
	 */
	new String:target_name[MAX_TARGET_LENGTH];
	new target_list[MAXPLAYERS], target_count;
	new bool:tn_is_ml;
 
	if ((target_count = ProcessTargetString(
			arg1,
			client,
			target_list,
			MAXPLAYERS,
			COMMAND_FILTER_ALIVE, /* Only allow alive players */
			target_name,
			sizeof(target_name),
			tn_is_ml)) <= 0)
	{
		/* This function replies to the admin with a failure message */
		ReplyToTargetError(client, target_count);
		return Plugin_Handled;
	}
 
	for (new i = 0; i < target_count; i++)
	{
		SlapPlayer(target_list[i], damage);
		LogAction(client, target_list[i], "\"%L\" slapped \"%L\" (damage %d)", client, target_list[i], damage);
	}
 
	if (tn_is_ml)
	{
		ShowActivity2(client, "[SM] ", "Slapped %t for %d damage!", target_name, damage);
	}
	else
	{
		ShowActivity2(client, "[SM] ", "Slapped %s for %d damage!", target_name, damage);
	}
 
	return Plugin_Handled;
}

Client and Entity Indexes

One major point of confusion with Half-Life 2 is the difference between the following things:

Client index
Entity index
Userid

The first answer is that clients are entities. Thus, a client index and an entity index are the same thing. When a SourceMod function asks for an entity index, a client index can be specified. When a SourceMod function asks for a client index, usually it means only a client index can be specified.

A fast way to check if an entity index is a client is checking whether it's between 1 and GetMaxClients() (inclusive). If a server has N client slots maximum, then entities 1 through N are always reserved for clients. Note that 0 is a valid entity index; it is the world entity (worldspawn).

A userid, on the other hand, is completely different. The server maintains a global "connection count" number, and it starts at 1. Each time a client connects, the connection count is incremented, and the client receives that new number as their userid.

For example, the first client to connect has a userid of 2. If he exits and rejoins, his userid will be 3 (unless another client joins in-between). Since clients are disconnected on mapchange, their userids change as well. Userids are a handy way to check if a client's connection status has changed.

SourceMod provides two functions for userids: GetClientOfUserId() and GetClientUserId().

Events

Events are informational notification messages passed between objects in the server. Many are also passed from the server to the client. They are defined in .res files under the hl2/resource folder and resource folders of specific mods. For a basic listing, see Source Game Events.

It is important to note a few concepts about events:

They are almost always informational. That is, blocking player_death will not stop a player from dying. It may block a HUD or console message or something else minor.
They almost always use userids instead of client indexes.
Just because it is in a resource file does not mean it is ever called, or works the way you expect it to. Mods are notorious at not properly documenting their event functionality.

An example of finding when a player dies:

public OnPluginStart()
{
   HookEvent("player_death", Event_PlayerDeath);
}
 
public Event_PlayerDeath(Handle:event, const String:name[], bool:dontBroadcast)
{
   new victim_id = GetEventInt(event, "userid");
   new attacker_id = GetEventInt(event, "attacker");
 
   new victim = GetClientOfUserId(victim_id);
   new attacker = GetClientOfUserId(attacker_id);
 
   /* CODE */
}

Callback Orders and Pairing

SourceMod has a number of builtin callbacks about the state of the server and plugin. Some of these are paired in special ways which is confusing to users.

Pairing

Pairing is SourceMod terminology. Examples of it are:

OnMapEnd() cannot be called without an OnMapStart(), and if OnMapStart() is called, it cannot be called again without an OnMapEnd().
OnClientConnected(N) for a given client N will only be called once, until an OnClientDisconnected(N) for the same client N is called (which is guaranteed to happen).

There is a formal definition of SourceMod's pairing. For two functions X and Y, both with input A, the following conditions hold:

If X is invoked with input A, it cannot be invoked again with the same input unless Y is called with input A.
If X is invoked with input A, it is guaranteed that Y will, at some point, be called with input A.
Y cannot be invoked with any input A unless X was called first with input A.
The relationship is described as, "X is paired with Y," and "Y is paired to X."

General Callbacks

These callbacks are listed in the order they are called, in the lifetime of a plugin and the server.

AskPluginLoad2() - Called once, immediately after the plugin is loaded from the disk. This function can be used to stop a plugin from loading and return a custom error message; return APLRes_Failure and use strcopy on to replace the error string. All CreateNative and RegPluginLibrary calls should be done here.
OnPluginStart() - Called once, after the plugin has been fully initialized and can proceed to load. Any run-time errors in this function will cause the plugin to fail to load. This is paired with OnPluginEnd().
OnMapStart() - Called every time the map loads. If the plugin is loaded late, and the map has already started, this function is called anyway after load, in order to preserve pairing. This function is paired with OnMapEnd().
OnConfigsExecuted() - Called once per map-change after servercfgfile (usually server.cfg), sourcemod.cfg, and all plugin config files have finished executing. If a plugin is loaded after this has happened, the callback is called anyway, in order to preserve pairing. This function is paired with OnMapEnd().
At this point, most game callbacks can occur, such as events and callbacks involving clients (or other things, like OnGameFrame).
OnMapEnd() - Called when the map is about to end. At this point, all clients are disconnected, but TIMER_NO_MAPCHANGE timers are not yet destroyed. This function is paired to OnMapStart().
OnPluginEnd() - Called once, immediately before the plugin is unloaded. This function is paired to OnPluginStart().

Client Callbacks

These callbacks are listed in no specific order, however, their documentation holds for both fake and real clients.

OnClientConnect() - Called when a player initiates a connection. You can block a player from connecting by returning Plugin_Stop and setting rejectmsg to an error message. This is paired with OnClientDisconnect() for successful connections only.
OnClientConnected(client) - Called after a player connects. Signifies that the player is in-game and IsClientConnected() will return true. This is paired with OnClientDisconnect() for successful connections only.
OnClientAuthorized() - Called when a player gets a Steam ID. It is important to note that this may never be called. It may occur any time in between ClientConnected and OnClientPreAdminCheck/OnClientDisconnect. Do not rely on it unless you are writing something that needs Steam IDs, and even then you should use OnClientPostAdminCheck().
OnClientPutInServer() - Signifies that the player is in-game and IsClientInGame() will return true.
Action:OnClientPreAdminCheck() - Called after the player is both authorized and in-game. Only used to override admin checks. For everything else, use OnClientPostAdminCheck.
OnClientPostAdminFilter() - Called after the player is both authorized and in-game. Used to override the permissions an admin has. For everything else, use OnClientPostAdminCheck.
OnClientPostAdminCheck() - Called after the player is both authorized and in-game. That is, both OnClientAuthorized() and OnClientPutInServer() have been invoked. This is the best callback for checking administrative access after connect.
OnClientDisconnect() - Called when a player's disconnection starts. This is paired to OnClientConnect().
OnClientDisconnect_Post() - Called when a player's disconnection ends. This is paired to OnClientConnect().

Frequently Asked Questions

Are plugins reloaded every mapchange?

Plugins, by default, are not reloaded on mapchange unless their timestamp changes. This is a feature so plugin authors have more flexibility with the state of their plugins.

Do I need to call CloseHandle in OnPluginEnd?

No. SourceMod automatically closes your Handles when your plugin is unloaded, in order to prevent memory errors.

Do I need to #include every individual .inc?

No. #include <sourcemod> will give you 95% of the .incs. Similarly, #include <sdktools> includes everything starting with <sdktools>.

Why don't some events fire?

There is no guarantee that events will fire. The event listing is not a specification, it is a list of the events that a game is capable of firing. Whether the game actually fires them is up to Valve or the developer.

Do I need to CloseHandle timers?

No. In fact, doing so may cause errors. Timers naturally die on their own unless they are infinite timers, in which case you can use KillTimer() or die gracefully by returning Plugin_Stop in the callback.

Are clients disconnected on mapchange?

All clients are fully disconnected before the map changes. They are all reconnected after the next map starts.

Introduction to SourceMod Plugins

Contents