diff --git a/README.md b/README.md
index 396b6a5..8177621 100644
--- a/README.md
+++ b/README.md
@@ -1,6 +1,10 @@
 # AI CLI
 
-AI agent in your terminal.
+AI assistant in your terminal.
+
+```
+$ ai "What are generator functions is JS?"
+```
 
 ## Installation
 
@@ -19,6 +23,7 @@ You will need a valid API key from one of currently supported AI inference provi
 Create a minimal `~/.airc` file (in your home directory):
 
 For Open AI:
+
 ```json
 {
   "providers": {
@@ -30,6 +35,7 @@ For Open AI:
 ```
 
 For Perplexity:
+
 ```json
 {
   "providers": {
@@ -63,6 +69,27 @@ ai: What do you call fake spaghetti? An impasta!
 me: _
 ```
 
+## CLI options
+
+CLI options are passed when invoking the `ai` commend:
+
+- `--help`: display available options
+- `--version`: display CLI version
+- `--interactive` (or `-i`): start an interactive session where user can ask follow-up questions
+- `--provider [name]` (or `-p [name]`): select an inference provider to use: `openai` or `perplexity` (alias `pplx`). You should have relevant API key in your `~/.airc` file.
+- `--model [name]` (or `-m [name]`): select a model to use. This should be a model available for selected provider.
+- `--verbose` (or `-V`): enable verbose mode
+
+## CLI commands
+
+CLI commands are available when using CLI in interactive mode.
+
+- `/help`: display available commands
+- `/exit`: exit the CLI
+- `/info`: display information about current session
+- `/verbose [on|off]`: enable/disable verbose mode
+- `/forget`: forget previous messages in the session
+
 ## Contributing
 
 See the [contributing guide](CONTRIBUTING.md) to learn how to contribute to the repository and the development workflow.
diff --git a/src/commands/prompt/commands.ts b/src/commands/prompt/commands.ts
new file mode 100644
index 0000000..7e5a7be
--- /dev/null
+++ b/src/commands/prompt/commands.ts
@@ -0,0 +1,56 @@
+import type { Message } from '../../inference';
+import * as output from '../../output';
+
+export interface CommandContext {
+  messages: Message[];
+  providerName: string;
+  config: {
+    model: string;
+    systemPrompt: string;
+  };
+}
+
+export function processCommand(input: string, context: CommandContext): boolean {
+  if (!input.startsWith('/')) {
+    return false;
+  }
+
+  const [command, ...args] = input.split(' ');
+  if (command === '/exit') {
+    process.exit(0);
+    // No need to return.
+  }
+
+  if (command === '/help') {
+    output.outputInfo('Available commands:');
+    output.outputInfo('- /exit: Exit the CLI');
+    output.outputInfo('- /info: Show current provider, model, and system prompt');
+    output.outputInfo('- /verbose [on|off]: Enable or disable verbose output');
+    output.outputInfo('- /forget: AI will forget previous messages');
+
+    return true;
+  }
+
+  if (command === '/info') {
+    output.outputInfo(`Provider: ${context.providerName}, model: ${context.config.model}`);
+    output.outputInfo('System prompt:', context.config.systemPrompt);
+    output.outputVerbose('Current context:', JSON.stringify(context.messages, null, 2));
+    return true;
+  }
+
+  if (command === '/verbose') {
+    output.setVerbose(args[0] !== 'off');
+    output.outputInfo(`Verbose mode: ${output.isVerbose() ? 'on' : 'off'}`);
+    return true;
+  }
+
+  if (command === '/forget') {
+    // Forget all messages from the context.
+    context.messages.length = 0;
+    output.outputInfo('AI will forget previous messages.', context.messages);
+    return true;
+  }
+
+  output.outputError(`Unknown command: ${command} ${args.join(' ')}`);
+  return true;
+}
diff --git a/src/commands/prompt.ts b/src/commands/prompt/index.ts
similarity index 82%
rename from src/commands/prompt.ts
rename to src/commands/prompt/index.ts
index 7d41f36..080e175 100644
--- a/src/commands/prompt.ts
+++ b/src/commands/prompt/index.ts
@@ -1,9 +1,10 @@
 import type { CommandModule } from 'yargs';
-import { parseConfigFile } from '../config-file';
-import { type Message } from '../inference';
-import { inputLine } from '../input';
-import * as output from '../output';
-import { providers, providerOptions, resolveProviderName } from '../providers';
+import { parseConfigFile } from '../../config-file';
+import { type Message } from '../../inference';
+import { inputLine } from '../../input';
+import * as output from '../../output';
+import { providers, providerOptions, resolveProviderName } from '../../providers';
+import { processCommand } from './commands';
 
 export interface PromptOptions {
   /** Interactive mode */
@@ -77,13 +78,13 @@ async function runInternal(initialPrompt: string, options: PromptOptions) {
     throw new Error(`Provider config not found: ${providerName}.`);
   }
 
-  const actualConfig = {
+  const config = {
     model: options.model ?? initialConfig.model,
     apiKey: initialConfig.apiKey,
     systemPrompt: initialConfig.systemPrompt,
   };
 
-  output.outputVerbose(`Using model: ${actualConfig.model}`);
+  output.outputVerbose(`Using model: ${config.model}`);
 
   const messages: Message[] = [];
 
@@ -92,27 +93,36 @@ async function runInternal(initialPrompt: string, options: PromptOptions) {
     output.outputAiProgress('Thinking...');
 
     messages.push({ role: 'user', content: initialPrompt });
-    const [content, response] = await provider.getChatCompletion(actualConfig, messages);
+    const [content, response] = await provider.getChatCompletion(config, messages);
 
     output.clearLine();
     output.outputVerbose(`Response: ${JSON.stringify(response, null, 2)}`);
     output.outputAi(content ?? '(null)');
     messages.push({ role: 'assistant', content: content ?? '' });
   } else {
-    output.outputAi('Hello, how can I help you? Press Ctrl+C to exit.');
+    output.outputAi('Hello, how can I help you? ');
   }
 
-  if (!options.interactive && initialPrompt) {
+  if (options.interactive || !initialPrompt) {
+    output.outputInfo(
+      'Type "/exit" or press Ctrl+C to exit. Type "/help" to see available commands.'
+    );
+  } else {
     process.exit(0);
   }
 
   // eslint-disable-next-line no-constant-condition
   while (true) {
     const userPrompt = await inputLine('me: ');
+    const isCommand = processCommand(userPrompt, { messages, providerName, config });
+    if (isCommand) {
+      continue;
+    }
+
     output.outputAiProgress('Thinking...');
 
     messages.push({ role: 'user', content: userPrompt });
-    const [content, response] = await provider.getChatCompletion(actualConfig, messages);
+    const [content, response] = await provider.getChatCompletion(config, messages);
     output.clearLine();
     output.outputVerbose(`Response Object: ${JSON.stringify(response, null, 2)}`);
     output.outputAi(content ?? '(null)');
diff --git a/src/output.ts b/src/output.ts
index dc20ae4..4837785 100644
--- a/src/output.ts
+++ b/src/output.ts
@@ -7,6 +7,10 @@ export function setVerbose(value: boolean) {
   verbose = value;
 }
 
+export function isVerbose() {
+  return verbose;
+}
+
 export function outputUser(message: string) {
   console.log('me:', message);
 }
@@ -22,12 +26,20 @@ export function outputAiProgress(message: string) {
 export function outputVerbose(message: string, ...args: unknown[]) {
   if (!verbose) return;
 
-  console.debug(chalk.grey(message), ...args);
+  console.debug(chalk.grey(message, ...args));
+}
+
+export function outputInfo(message: string, ...args: unknown[]) {
+  console.log(chalk.dim(message, ...args));
 }
 
 export function outputError(error: unknown, ...args: unknown[]) {
   const message = extractErrorMessage(error);
-  console.error(chalk.red(`ERROR: ${message}`), error, ...args);
+  if (error === message) {
+    console.error(chalk.red(`ERROR: ${message}`, ...args));
+  } else {
+    console.error(chalk.red(`ERROR: ${message}`, error, ...args));
+  }
 }
 
 /**