diff --git a/CLAUDE.md b/CLAUDE.md
index 33094d46..ed48143b 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -554,7 +554,7 @@ Typed arrays: `let nums: array<i32> = [1, 2, 3];`
 
 ---
 
-## Standard Library (42 modules)
+## Standard Library (43 modules)
 
 Import with `@stdlib/` prefix:
 ```hemlock
@@ -592,6 +592,7 @@ import { TcpStream, UdpSocket } from "@stdlib/net";
 | `os` | platform, arch, cpu_count, hostname |
 | `path` | File path manipulation |
 | `process` | fork, exec, wait, kill |
+| `prompt` | LLM prompt templating, chat formats, conversation builder |
 | `random` | Random number generation |
 | `regex` | compile, test (POSIX ERE) |
 | `retry` | Retry logic with backoff |
@@ -979,7 +980,7 @@ make parity
 - Manual memory management with `talloc()` and `sizeof()`
 - Async/await with true pthread parallelism
 - Atomic operations for lock-free concurrent programming
-- 42 stdlib modules (+ arena, assert, semver, toml, retry, iter, random, shell, termios, vector)
+- 43 stdlib modules (+ arena, assert, semver, toml, retry, iter, random, shell, termios, vector, prompt)
 - FFI for C interop with `export extern fn` for reusable library wrappers
 - FFI struct support in compiler (pass C structs by value)
 - FFI pointer helpers (`ptr_null`, `ptr_read_*`, `ptr_write_*`)
diff --git a/stdlib/docs/prompt.md b/stdlib/docs/prompt.md
new file mode 100644
index 00000000..afbd0746
--- /dev/null
+++ b/stdlib/docs/prompt.md
@@ -0,0 +1,329 @@
+# @stdlib/prompt - LLM Prompt Templating
+
+## Overview
+
+The prompt module provides utilities for building multi-turn LLM conversations, applying chat templates, and managing message arrays. It uses `@stdlib/jinja` under the hood for template rendering.
+
+## Quick Start
+
+```hemlock
+import { system, user, format_chat } from "@stdlib/prompt";
+
+let msgs = [
+    system("You are a helpful assistant."),
+    user("What is 2 + 2?")
+];
+
+let prompt = format_chat(msgs, "chatml");
+print(prompt);
+// <|im_start|>system
+// You are a helpful assistant.<|im_end|>
+// <|im_start|>user
+// What is 2 + 2?<|im_end|>
+// <|im_start|>assistant
+```
+
+## API Reference
+
+### Message Constructors
+
+#### `message(role: string, content: string): object`
+
+Create a message with any role.
+
+```hemlock
+let msg = message("system", "You are helpful.");
+print(msg.role);     // "system"
+print(msg.content);  // "You are helpful."
+```
+
+#### `system(content: string): object`
+
+Create a system message.
+
+```hemlock
+let msg = system("You are a coding assistant.");
+```
+
+#### `user(content: string): object`
+
+Create a user message.
+
+```hemlock
+let msg = user("Write a function to add two numbers.");
+```
+
+#### `assistant(content: string): object`
+
+Create an assistant message.
+
+```hemlock
+let msg = assistant("Here's a function that adds two numbers...");
+```
+
+#### `tool(content: string, name?: string): object`
+
+Create a tool/function result message with optional tool name.
+
+```hemlock
+let msg = tool("42", "calculator");
+print(msg.role);    // "tool"
+print(msg.name);    // "calculator"
+```
+
+### Template Formatting
+
+#### `format_chat(messages: array, template_name: string): string`
+
+Format a message array with a named built-in template.
+
+**Parameters:**
+- `messages` - Array of message objects (each with `.role` and `.content`)
+- `template_name` - One of: `chatml`, `llama3`, `alpaca`, `vicuna`, `zephyr`, `gemma`, `plain`
+
+**Returns:** Formatted prompt string
+
+```hemlock
+import { system, user, assistant, format_chat } from "@stdlib/prompt";
+
+let msgs = [
+    system("You are helpful."),
+    user("Hi!"),
+    assistant("Hello!"),
+    user("How are you?")
+];
+
+// ChatML format
+print(format_chat(msgs, "chatml"));
+
+// Llama 3 format
+print(format_chat(msgs, "llama3"));
+
+// Plain text format
+print(format_chat(msgs, "plain"));
+```
+
+#### `apply_template(template: string, messages: array, context?: object): string`
+
+Apply a custom Jinja template to a message array. The `messages` variable is automatically available in the template. Additional variables can be passed via `context`.
+
+```hemlock
+import { system, user, apply_template } from "@stdlib/prompt";
+
+let tmpl = "{% for msg in messages %}[{{ msg.role|upper }}] {{ msg.content }}\n{% endfor %}";
+let msgs = [system("Be brief."), user("Hello!")];
+
+print(apply_template(tmpl, msgs));
+// [SYSTEM] Be brief.
+// [USER] Hello!
+
+// With extra context
+let tmpl2 = "Model: {{ model }}\n{% for msg in messages %}{{ msg.content }}\n{% endfor %}";
+print(apply_template(tmpl2, msgs, { model: "gpt-4" }));
+```
+
+### Conversation Builder
+
+#### `conversation(template_name?: string): object`
+
+Create a conversation builder for incremental message construction. Defaults to `"chatml"` template.
+
+**Methods:**
+- `add_system(content)` - Add a system message (returns self for chaining)
+- `add_user(content)` - Add a user message (returns self for chaining)
+- `add_assistant(content)` - Add an assistant message (returns self for chaining)
+- `add_message(role, content)` - Add a message with any role (returns self for chaining)
+- `format()` - Format all messages using the configured template
+- `format_with(template)` - Format using a custom Jinja template
+- `count()` - Get the number of messages
+- `last_message()` - Get the last message object (or null)
+- `reset()` - Remove all messages (returns self for chaining)
+- `by_role(role)` - Get messages filtered by role
+
+```hemlock
+import { conversation } from "@stdlib/prompt";
+
+let conv = conversation("chatml");
+conv.add_system("You are a math tutor.");
+conv.add_user("What is calculus?");
+conv.add_assistant("Calculus is the study of continuous change...");
+conv.add_user("Can you give an example?");
+
+print(conv.count());          // 4
+print(conv.last_message().role);  // "user"
+print(conv.format());         // Full ChatML-formatted prompt
+
+// Filter by role
+let user_msgs = conv.by_role("user");
+print(user_msgs.length);      // 2
+```
+
+### Utility Functions
+
+#### `templates(): array`
+
+List all available built-in template names.
+
+```hemlock
+import { templates } from "@stdlib/prompt";
+
+for (name in templates()) {
+    print(name);
+}
+// chatml, llama3, alpaca, vicuna, zephyr, gemma, plain
+```
+
+#### `get_template_source(name: string): string`
+
+Get the raw Jinja template string for a built-in template. Useful for inspecting or modifying templates.
+
+```hemlock
+import { get_template_source } from "@stdlib/prompt";
+
+let tmpl = get_template_source("chatml");
+print(tmpl);
+```
+
+## Built-in Templates
+
+### ChatML (`chatml`)
+
+Used by OpenAI, Qwen, Yi, and many other models.
+
+```
+<|im_start|>system
+{content}<|im_end|>
+<|im_start|>user
+{content}<|im_end|>
+<|im_start|>assistant
+```
+
+### Llama 3 (`llama3`)
+
+Used by Meta's Llama 3 and Llama 3.1 models.
+
+```
+<|start_header_id|>system<|end_header_id|>
+
+{content}<|eot_id|><|start_header_id|>user<|end_header_id|>
+
+{content}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
+
+```
+
+### Alpaca (`alpaca`)
+
+Instruction-following format used by Stanford Alpaca.
+
+```
+### Instruction:
+{system content}
+
+### Input:
+{user content}
+
+### Response:
+```
+
+### Vicuna (`vicuna`)
+
+Used by Vicuna/LMSys models.
+
+```
+{system content}
+
+USER: {user content}
+ASSISTANT:
+```
+
+### Zephyr (`zephyr`)
+
+Used by HuggingFace Zephyr (Mistral-based).
+
+```
+<|system|>
+{content}<|endoftext|>
+<|user|>
+{content}<|endoftext|>
+<|assistant|>
+```
+
+### Gemma (`gemma`)
+
+Used by Google Gemma models. System messages are formatted as user turns.
+
+```
+<start_of_turn>user
+{content}<end_of_turn>
+<start_of_turn>model
+```
+
+### Plain (`plain`)
+
+Simple text format with no special tokens, useful for debugging.
+
+```
+[system]: {content}
+[user]: {content}
+[assistant]:
+```
+
+## Examples
+
+### Multi-turn Conversation
+
+```hemlock
+import { conversation } from "@stdlib/prompt";
+
+let conv = conversation("chatml");
+conv.add_system("You are a Python expert.");
+
+// Simulate a multi-turn conversation
+conv.add_user("How do I read a file in Python?");
+conv.add_assistant("Use open() with a context manager:\n```python\nwith open('file.txt') as f:\n    content = f.read()\n```");
+conv.add_user("What about binary files?");
+
+print(conv.format());
+```
+
+### Custom Template for Tool Use
+
+```hemlock
+import { system, user, tool, apply_template } from "@stdlib/prompt";
+
+let tmpl = `{%- for msg in messages -%}
+{%- if msg.role == 'system' -%}
+<|system|>{{ msg.content }}</s>
+{%- elif msg.role == 'user' -%}
+<|user|>{{ msg.content }}</s>
+{%- elif msg.role == 'tool' -%}
+<|tool|>[{{ msg.name }}] {{ msg.content }}</s>
+{%- endif -%}
+{%- endfor -%}
+<|assistant|>`;
+
+let msgs = [
+    system("You can use tools."),
+    user("What is 6 * 7?"),
+    tool("42", "calculator")
+];
+
+print(apply_template(tmpl, msgs));
+```
+
+### Comparing Template Formats
+
+```hemlock
+import { system, user, format_chat, templates } from "@stdlib/prompt";
+
+let msgs = [
+    system("You are helpful."),
+    user("Hello!")
+];
+
+for (name in templates()) {
+    print("=== " + name + " ===");
+    print(format_chat(msgs, name));
+    print("");
+}
+```
diff --git a/stdlib/prompt.hml b/stdlib/prompt.hml
new file mode 100644
index 00000000..8ceffaa7
--- /dev/null
+++ b/stdlib/prompt.hml
@@ -0,0 +1,244 @@
+// @stdlib/prompt - LLM prompt templating and conversation management
+//
+// Provides utilities for building multi-turn LLM conversations,
+// applying chat templates, and managing system/user/assistant messages.
+//
+// Usage:
+// import { message, system, user, assistant, format_chat, apply_template } from "@stdlib/prompt";
+//
+// let msgs = [system("You are helpful."), user("Hello!")];
+// let prompt = format_chat(msgs, "chatml");
+// print(prompt);
+//
+// Built-in templates: chatml, llama3, alpaca, vicuna, zephyr, gemma, plain
+//
+// Custom templates use Jinja syntax via @stdlib/jinja.
+
+import { render } from "@stdlib/jinja";
+
+// ============================================================================
+// Message Constructors
+// ============================================================================
+
+// Create a message with a given role and content
+export fn message(role, content): object {
+	if (typeof(role) != "string") {
+		throw "message() role must be a string";
+	}
+	if (typeof(content) != "string") {
+		content = "" + content;
+	}
+	return { role: role, content: content };
+}
+
+// Create a system message
+export fn system(content): object {
+	return message("system", content);
+}
+
+// Create a user message
+export fn user(content): object {
+	return message("user", content);
+}
+
+// Create an assistant message
+export fn assistant(content): object {
+	return message("assistant", content);
+}
+
+// Create a tool message with optional tool name
+export fn tool(content, name?: null): object {
+	let msg = message("tool", content);
+	if (name != null) {
+		msg["name"] = name;
+	}
+	return msg;
+}
+
+// ============================================================================
+// Built-in Chat Templates (Jinja syntax)
+// ============================================================================
+
+// ChatML format (OpenAI, Qwen, Yi, etc.)
+let CHATML = "{% for msg in messages %}<|im_start|>{{ msg.role }}\n{{ msg.content }}<|im_end|>\n{% endfor %}<|im_start|>assistant\n";
+
+// Llama 3 / Llama 3.1 format
+let LLAMA3 = "{% for msg in messages %}<|start_header_id|>{{ msg.role }}<|end_header_id|>\n\n{{ msg.content }}<|eot_id|>{% endfor %}<|start_header_id|>assistant<|end_header_id|>\n\n";
+
+// Alpaca format (instruction-following)
+let ALPACA_TMPL = "{% for msg in messages %}{% if msg.role == 'system' %}### Instruction:\n{{ msg.content }}\n\n{% elif msg.role == 'user' %}### Input:\n{{ msg.content }}\n\n{% elif msg.role == 'assistant' %}### Response:\n{{ msg.content }}\n\n{% endif %}{% endfor %}### Response:\n";
+
+// Vicuna format
+let VICUNA = "{% for msg in messages %}{% if msg.role == 'system' %}{{ msg.content }}\n\n{% elif msg.role == 'user' %}USER: {{ msg.content }}\n{% elif msg.role == 'assistant' %}ASSISTANT: {{ msg.content }}\n{% endif %}{% endfor %}ASSISTANT:";
+
+// Zephyr format (Mistral)
+let ZEPHYR = "{% for msg in messages %}<|{{ msg.role }}|>\n{{ msg.content }}<|endoftext|>\n{% endfor %}<|assistant|>\n";
+
+// Gemma format
+let GEMMA = "{% for msg in messages %}{% if msg.role == 'user' %}<start_of_turn>user\n{{ msg.content }}<end_of_turn>\n{% elif msg.role == 'assistant' %}<start_of_turn>model\n{{ msg.content }}<end_of_turn>\n{% elif msg.role == 'system' %}<start_of_turn>user\n{{ msg.content }}<end_of_turn>\n{% endif %}{% endfor %}<start_of_turn>model\n";
+
+// Plain text format (no special tokens)
+let PLAIN = "{% for msg in messages %}[{{ msg.role }}]: {{ msg.content }}\n{% endfor %}[assistant]:";
+
+fn get_template(name): string {
+	if (name == "chatml") {
+		return CHATML;
+	}
+	if (name == "llama3") {
+		return LLAMA3;
+	}
+	if (name == "alpaca") {
+		return ALPACA_TMPL;
+	}
+	if (name == "vicuna") {
+		return VICUNA;
+	}
+	if (name == "zephyr") {
+		return ZEPHYR;
+	}
+	if (name == "gemma") {
+		return GEMMA;
+	}
+	if (name == "plain") {
+		return PLAIN;
+	}
+	throw "Unknown chat template: " + name + ". Available: chatml, llama3, alpaca, vicuna, zephyr, gemma, plain";
+}
+
+// ============================================================================
+// Template Formatting
+// ============================================================================
+
+// Format a message array with a named built-in template
+// Parameters:
+//   messages: array of message objects (each with .role and .content)
+//   template_name: string - one of: chatml, llama3, alpaca, vicuna, zephyr, gemma, plain
+// Returns: formatted prompt string
+export fn format_chat(messages, template_name): string {
+	if (typeof(messages) != "array") {
+		throw "format_chat() messages must be an array";
+	}
+	if (typeof(template_name) != "string") {
+		throw "format_chat() template_name must be a string";
+	}
+	let tmpl = get_template(template_name);
+	return render(tmpl, { messages: messages });
+}
+
+// Apply a custom Jinja template to a message array
+// Parameters:
+//   template: string - Jinja template string
+//   messages: array of message objects
+//   context: object - additional variables available in the template (optional)
+// Returns: rendered string
+export fn apply_template(template, messages, context?: null): string {
+	if (typeof(template) != "string") {
+		throw "apply_template() template must be a string";
+	}
+	if (typeof(messages) != "array") {
+		throw "apply_template() messages must be an array";
+	}
+	let ctx = {};
+	if (context != null) {
+		if (typeof(context) != "object") {
+			throw "apply_template() context must be an object";
+		}
+		let keys = context.keys();
+		for (key in keys) {
+			ctx[key] = context[key];
+		}
+	}
+	ctx["messages"] = messages;
+	return render(template, ctx);
+}
+
+// ============================================================================
+// Conversation Builder
+// ============================================================================
+
+// Create a conversation builder for incremental message construction
+// Parameters:
+//   template_name: string - built-in template name (default: "chatml")
+// Returns: conversation object with methods
+export fn conversation(template_name?: "chatml"): object {
+	let msgs: array = [];
+	let tmpl_name = template_name;
+
+	let conv = {
+		messages: msgs,
+		template_name: tmpl_name
+	};
+
+	// Add methods after conv is declared so closures can reference it
+	conv["add_system"] = fn(content) {
+		msgs.push(system(content));
+		return conv;
+	};
+
+	conv["add_user"] = fn(content) {
+		msgs.push(user(content));
+		return conv;
+	};
+
+	conv["add_assistant"] = fn(content) {
+		msgs.push(assistant(content));
+		return conv;
+	};
+
+	conv["add_message"] = fn(role, content) {
+		msgs.push(message(role, content));
+		return conv;
+	};
+
+	conv["format"] = fn(): string {
+		return format_chat(msgs, tmpl_name);
+	};
+
+	conv["format_with"] = fn(custom_template): string {
+		return apply_template(custom_template, msgs);
+	};
+
+	conv["count"] = fn(): i32 {
+		return msgs.length;
+	};
+
+	conv["last_message"] = fn() {
+		if (msgs.length == 0) {
+			return null;
+		}
+		return msgs[msgs.length - 1];
+	};
+
+	conv["reset"] = fn() {
+		while (msgs.length > 0) {
+			msgs.pop();
+		}
+		return conv;
+	};
+
+	conv["by_role"] = fn(role): array {
+		let filtered: array = [];
+		for (msg in msgs) {
+			if (msg.role == role) {
+				filtered.push(msg);
+			}
+		}
+		return filtered;
+	};
+
+	return conv;
+}
+
+// ============================================================================
+// Utility Functions
+// ============================================================================
+
+// List all available built-in template names
+export fn templates(): array {
+	return ["chatml", "llama3", "alpaca", "vicuna", "zephyr", "gemma", "plain"];
+}
+
+// Get the raw Jinja template string for a built-in template
+export fn get_template_source(name): string {
+	return get_template(name);
+}
diff --git a/tests/parity/modules/stdlib_prompt.expected b/tests/parity/modules/stdlib_prompt.expected
new file mode 100644
index 00000000..35b76b5b
--- /dev/null
+++ b/tests/parity/modules/stdlib_prompt.expected
@@ -0,0 +1,44 @@
+system
+You are helpful.
+user
+Hello!
+assistant
+Hi there!
+custom
+test
+tool
+result
+calc
+[system]: Be brief.
+[user]: Hi!
+[assistant]:
+<|im_start|>system
+Be brief.<|im_end|>
+<|im_start|>user
+Hi!<|im_end|>
+<|im_start|>assistant
+
+(system) Be brief.
+(user) Hi!
+
+Model: test-model
+Be brief.
+Hi!
+
+4
+user
+Follow-up?
+2
+[system]: System prompt.
+[user]: Question?
+[assistant]: Answer.
+[user]: Follow-up?
+[assistant]:
+0
+7
+chatml
+plain
+true
+tool
+data
+done
diff --git a/tests/parity/modules/stdlib_prompt.hml b/tests/parity/modules/stdlib_prompt.hml
new file mode 100644
index 00000000..480340bf
--- /dev/null
+++ b/tests/parity/modules/stdlib_prompt.hml
@@ -0,0 +1,93 @@
+// Test @stdlib/prompt module
+
+import { message, system, user, assistant, tool, format_chat, apply_template, conversation, templates, get_template_source } from "@stdlib/prompt";
+
+// === Message constructors ===
+
+let sys = system("You are helpful.");
+print(sys.role);
+print(sys.content);
+
+let usr = user("Hello!");
+print(usr.role);
+print(usr.content);
+
+let ast = assistant("Hi there!");
+print(ast.role);
+print(ast.content);
+
+let msg = message("custom", "test");
+print(msg.role);
+print(msg.content);
+
+let tl = tool("result", "calc");
+print(tl.role);
+print(tl.content);
+print(tl.name);
+
+// === format_chat with plain template ===
+
+let msgs = [system("Be brief."), user("Hi!")];
+let plain = format_chat(msgs, "plain");
+print(plain);
+
+// === format_chat with chatml ===
+
+let chatml = format_chat(msgs, "chatml");
+print(chatml);
+
+// === apply_template with custom template ===
+
+let custom_tmpl = "{% for msg in messages %}({{ msg.role }}) {{ msg.content }}\n{% endfor %}";
+let custom = apply_template(custom_tmpl, msgs);
+print(custom);
+
+// === apply_template with extra context ===
+
+let ctx_tmpl = "Model: {{ model_name }}\n{% for msg in messages %}{{ msg.content }}\n{% endfor %}";
+let ctx_result = apply_template(ctx_tmpl, msgs, { model_name: "test-model" });
+print(ctx_result);
+
+// === conversation builder ===
+
+let conv = conversation("plain");
+conv.add_system("System prompt.");
+conv.add_user("Question?");
+conv.add_assistant("Answer.");
+conv.add_user("Follow-up?");
+
+print(conv.count());
+print(conv.last_message().role);
+print(conv.last_message().content);
+
+// by_role
+let user_msgs = conv.by_role("user");
+print(user_msgs.length);
+
+// format
+let formatted = conv.format();
+print(formatted);
+
+// reset
+conv.reset();
+print(conv.count());
+
+// === templates list ===
+
+let tmpl_list = templates();
+print(tmpl_list.length);
+print(tmpl_list[0]);
+print(tmpl_list[6]);
+
+// === get_template_source ===
+
+let src = get_template_source("plain");
+print(src.contains("msg.role"));
+
+// === tool message without name ===
+
+let tl2 = tool("data");
+print(tl2.role);
+print(tl2.content);
+
+print("done");