feat: add /usage command to show model info and token usage

Adds token usage tracking to AgentInstance using atomic counters
(TotalPromptTokens, TotalCompletionTokens, TotalRequests) that
accumulate after each LLM call. The /usage command displays current
model name, max tokens, temperature, and session token statistics.
Available in all interactive modes like /help.

https://claude.ai/code/session_01PoHBfe7eKmhHvVF12W3xZ2

cmd/picoclaw/cmd_agent.go

@@ -109,6 +109,7 @@ func agentCmd() {
 	} else {
 		fmt.Printf("%s Interactive mode (Ctrl+C to exit)\n", logo)
 		fmt.Println("  /help    - show detailed help")
+		fmt.Println("  /usage   - show model info and token usage")
 		fmt.Println("  /cmd     - switch to command mode")
 		fmt.Println("  /pico    - switch to chat mode")
 		fmt.Println("  /hipico  - AI assistance in command mode")
@@ -163,11 +164,15 @@ func interactiveMode(agentLoop *agent.AgentLoop, sessionKey string) {
 			return
 		}
 
-		// /help works in all modes
+		// /help and /usage work in all modes
 		if input == "/help" {
 			printHelp()
 			continue
 		}
+		if input == "/usage" {
+			printUsage(agentLoop)
+			continue
+		}
 
 		switch mode {
 		case modePico:
@@ -284,11 +289,15 @@ func simpleInteractiveMode(agentLoop *agent.AgentLoop, sessionKey string) {
 			return
 		}
 
-		// /help works in all modes
+		// /help and /usage work in all modes
 		if input == "/help" {
 			printHelp()
 			continue
 		}
+		if input == "/usage" {
+			printUsage(agentLoop)
+			continue
+		}
 
 		switch mode {
 		case modePico:
@@ -383,6 +392,7 @@ PicoClaw has three interactive modes:
 
 Commands (available in all modes):
   /help      Show this help message
+  /usage     Show model info and token usage
   exit       Exit PicoClaw
   quit       Exit PicoClaw
   Ctrl+C     Exit PicoClaw
@@ -410,6 +420,35 @@ Examples:
 `, logo, logo, logo, logo)
 }
 
+// printUsage displays current model information and accumulated token usage.
+func printUsage(agentLoop *agent.AgentLoop) {
+	info := agentLoop.GetUsageInfo()
+	if info == nil {
+		fmt.Println("No usage information available.")
+		return
+	}
+	fmt.Printf(`%s Usage
+━━━━━━━━━━━━━━━━━━━━━━
+Model:              %s
+Max tokens:         %d
+Temperature:        %.1f
+
+Token usage (this session):
+  Prompt tokens:    %d
+  Completion tokens:%d
+  Total tokens:     %d
+  Requests:         %d
+`, logo,
+		info["model"],
+		info["max_tokens"],
+		info["temperature"],
+		info["prompt_tokens"],
+		info["completion_tokens"],
+		info["total_tokens"],
+		info["requests"],
+	)
+}
+
 // executeShellCommand runs a shell command in the current working directory
 // and prints the output. It also handles the cd command to change directories.
 func executeShellCommand(input string) {

pkg/agent/instance.go

@@ -4,6 +4,7 @@ import (
 	"os"
 	"path/filepath"
 	"strings"
+	"sync/atomic"
 
 	"github.com/sipeed/picoclaw/pkg/config"
 	"github.com/sipeed/picoclaw/pkg/providers"
@@ -31,6 +32,21 @@ type AgentInstance struct {
 	Subagents      *config.SubagentsConfig
 	SkillsFilter   []string
 	Candidates     []providers.FallbackCandidate
+
+	// Accumulated token usage counters (atomic for concurrent safety)
+	TotalPromptTokens     atomic.Int64
+	TotalCompletionTokens atomic.Int64
+	TotalRequests         atomic.Int64
+}
+
+// AddUsage accumulates token usage from a single LLM response.
+func (a *AgentInstance) AddUsage(usage *providers.UsageInfo) {
+	if usage == nil {
+		return
+	}
+	a.TotalPromptTokens.Add(int64(usage.PromptTokens))
+	a.TotalCompletionTokens.Add(int64(usage.CompletionTokens))
+	a.TotalRequests.Add(1)
 }
 
 // NewAgentInstance creates an agent instance from config.

pkg/agent/loop.go

@@ -594,6 +594,9 @@ func (al *AgentLoop) runLLMIteration(
 			return "", iteration, fmt.Errorf("LLM call failed after retries: %w", err)
 		}
 
+		// Accumulate token usage
+		agent.AddUsage(response.Usage)
+
 		// Check if no tool calls - we're done
 		if len(response.ToolCalls) == 0 {
 			finalContent = response.Content
@@ -852,6 +855,25 @@ func (al *AgentLoop) GetStartupInfo() map[string]any {
 	return info
 }
 
+// GetUsageInfo returns accumulated token usage for the default agent.
+func (al *AgentLoop) GetUsageInfo() map[string]any {
+	agent := al.registry.GetDefaultAgent()
+	if agent == nil {
+		return nil
+	}
+	promptTokens := agent.TotalPromptTokens.Load()
+	completionTokens := agent.TotalCompletionTokens.Load()
+	return map[string]any{
+		"model":             agent.Model,
+		"max_tokens":        agent.MaxTokens,
+		"temperature":       agent.Temperature,
+		"prompt_tokens":     promptTokens,
+		"completion_tokens": completionTokens,
+		"total_tokens":      promptTokens + completionTokens,
+		"requests":          agent.TotalRequests.Load(),
+	}
+}
+
 // formatMessagesForLog formats messages for logging
 func formatMessagesForLog(messages []providers.Message) string {
 	if len(messages) == 0 {