diff --git a/cmd/gonzo/main.go b/cmd/gonzo/main.go
index c924626..f264910 100644
--- a/cmd/gonzo/main.go
+++ b/cmd/gonzo/main.go
@@ -165,7 +165,7 @@ func init() {
 	rootCmd.Flags().IntP("log-buffer", "b", 1000, "Maximum log buffer size")
 	rootCmd.Flags().BoolP("test-mode", "t", false, "Run in test mode (works without TTY)")
 	rootCmd.Flags().BoolP("version", "v", false, "Print version information")
-	rootCmd.Flags().String("ai-provider", "", "AI provider to use: 'openai' (default, requires OPENAI_API_KEY), 'claude-code' (requires claude CLI)")
+	rootCmd.Flags().String("ai-provider", "", "AI provider to use: 'openai' (default, requires OPENAI_API_KEY), 'claude-code' (requires claude CLI), 'litellm' (requires LITELLM_API_KEY)")
 	rootCmd.Flags().String("ai-model", "", "AI model to use for log analysis (auto-selects best available if not specified)")
 	rootCmd.Flags().StringSliceP("file", "f", []string{}, "Files or file globs to read logs from (can specify multiple)")
 	rootCmd.Flags().Bool("follow", false, "Follow log files like 'tail -f' (watch for new lines in real-time)")
diff --git a/internal/ai/factory.go b/internal/ai/factory.go
index a6cd580..c68e311 100644
--- a/internal/ai/factory.go
+++ b/internal/ai/factory.go
@@ -12,6 +12,10 @@ const (
 	// ProviderClaudeCode uses the Claude Code CLI
 	ProviderClaudeCode ProviderType = "claude-code"
 
+	// ProviderLiteLLM uses a LiteLLM proxy (OpenAI-wire compatible gateway
+	// to 100+ providers). Requires LITELLM_API_KEY.
+	ProviderLiteLLM ProviderType = "litellm"
+
 	// ProviderAuto auto-detects the provider based on environment.
 	// This is the default and maintains backwards compatibility:
 	// - If OPENAI_API_KEY is set, uses OpenAI client
@@ -43,6 +47,11 @@ func NewClient(provider ProviderType, model string) (Client, error) {
 		// so TUI can display the validation error
 		return NewClaudeCodeClient(model), nil
 
+	case ProviderLiteLLM:
+		// Explicitly requested LiteLLM - returns client even if not validated
+		// so TUI can display the validation error
+		return NewLiteLLMClient(model), nil
+
 	case ProviderAuto:
 		// Auto-detect: maintain backwards compatibility
 		// Only try OpenAI (the previous default behavior)
@@ -60,5 +69,5 @@ func NewClient(provider ProviderType, model string) (Client, error) {
 
 // ValidProviders returns the list of valid provider type strings
 func ValidProviders() []string {
-	return []string{string(ProviderOpenAI), string(ProviderClaudeCode)}
+	return []string{string(ProviderOpenAI), string(ProviderClaudeCode), string(ProviderLiteLLM)}
 }
diff --git a/internal/ai/litellm.go b/internal/ai/litellm.go
new file mode 100644
index 0000000..91edf18
--- /dev/null
+++ b/internal/ai/litellm.go
@@ -0,0 +1,54 @@
+package ai
+
+import (
+	"net/http"
+	"os"
+	"time"
+)
+
+// LiteLLM proxy defaults. A LiteLLM proxy speaks the OpenAI wire format, so the
+// LiteLLM provider reuses the tested OpenAIClient and only differs in its
+// defaults and environment variables. Point it at your proxy to reach 100+
+// providers (OpenAI, Anthropic, Gemini, Bedrock, Vertex AI, Azure, ...) through
+// a single endpoint with unified auth, routing and observability.
+const defaultLiteLLMBaseURL = "http://localhost:4000/v1"
+
+// NewLiteLLMClient creates a client for a LiteLLM proxy.
+//
+// Configuration (mirrors the OpenAI provider, with LiteLLM-specific env vars):
+//   - LITELLM_API_KEY:  the proxy master/virtual key (required)
+//   - LITELLM_API_BASE: the proxy base URL (default http://localhost:4000/v1)
+//
+// LiteLLM is OpenAI-wire compatible, so this reuses OpenAIClient (and therefore
+// its chat, model-discovery, validation and retry behavior) verbatim.
+func NewLiteLLMClient(model string) *OpenAIClient {
+	apiKey := os.Getenv("LITELLM_API_KEY")
+	if apiKey == "" {
+		return &OpenAIClient{
+			Validated:     false,
+			ValidationErr: "LITELLM_API_KEY environment variable not set",
+			ServiceName:   "LiteLLM",
+		}
+	}
+
+	baseURL := os.Getenv("LITELLM_API_BASE")
+	if baseURL == "" {
+		baseURL = defaultLiteLLMBaseURL
+	}
+
+	client := &OpenAIClient{
+		APIKey:          apiKey,
+		BaseURL:         baseURL,
+		Model:           model,
+		ServiceName:     "LiteLLM",
+		AutoSelectModel: model == "",
+		HTTPClient: &http.Client{
+			Timeout: 60 * time.Second,
+		},
+	}
+
+	// Validate configuration and get available models (via the proxy's /models).
+	client.ValidateConfiguration()
+
+	return client
+}
diff --git a/internal/ai/litellm_test.go b/internal/ai/litellm_test.go
new file mode 100644
index 0000000..fd2f6f0
--- /dev/null
+++ b/internal/ai/litellm_test.go
@@ -0,0 +1,77 @@
+package ai
+
+import (
+	"slices"
+	"testing"
+)
+
+func TestNewLiteLLMClient_NoAPIKeyReturnsUnvalidated(t *testing.T) {
+	t.Setenv("LITELLM_API_KEY", "")
+
+	client := NewLiteLLMClient("")
+
+	if client.Validated {
+		t.Fatal("expected unvalidated client when LITELLM_API_KEY is unset")
+	}
+	if client.ServiceName != "LiteLLM" {
+		t.Fatalf("expected ServiceName %q, got %q", "LiteLLM", client.ServiceName)
+	}
+	if client.ValidationErr == "" {
+		t.Fatal("expected a validation error message when LITELLM_API_KEY is unset")
+	}
+}
+
+func TestNewLiteLLMClient_DefaultBaseURL(t *testing.T) {
+	// Set a key so the constructor proceeds past the no-key guard. Validation
+	// makes a network call and may fail (no proxy running), but BaseURL/defaults
+	// are assigned before that.
+	t.Setenv("LITELLM_API_KEY", "sk-test")
+	t.Setenv("LITELLM_API_BASE", "")
+
+	client := NewLiteLLMClient("some-model")
+
+	if client.BaseURL != defaultLiteLLMBaseURL {
+		t.Fatalf("expected default BaseURL %q, got %q", defaultLiteLLMBaseURL, client.BaseURL)
+	}
+	if client.APIKey != "sk-test" {
+		t.Fatalf("expected APIKey to be forwarded, got %q", client.APIKey)
+	}
+	if client.ServiceName != "LiteLLM" {
+		t.Fatalf("expected ServiceName %q, got %q", "LiteLLM", client.ServiceName)
+	}
+	if client.Model != "some-model" {
+		t.Fatalf("expected Model %q, got %q", "some-model", client.Model)
+	}
+}
+
+func TestNewLiteLLMClient_CustomBaseURL(t *testing.T) {
+	t.Setenv("LITELLM_API_KEY", "sk-test")
+	t.Setenv("LITELLM_API_BASE", "http://127.0.0.1:4000/v1")
+
+	client := NewLiteLLMClient("m")
+
+	if client.BaseURL != "http://127.0.0.1:4000/v1" {
+		t.Fatalf("expected custom BaseURL to be honored, got %q", client.BaseURL)
+	}
+}
+
+func TestNewClient_LiteLLMProvider(t *testing.T) {
+	t.Setenv("LITELLM_API_KEY", "")
+
+	client, err := NewClient(ProviderLiteLLM, "")
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if client == nil {
+		t.Fatal("expected a non-nil client for explicit litellm provider")
+	}
+	if got := client.GetValidationStatus().ServiceName; got != "LiteLLM" {
+		t.Fatalf("expected ServiceName %q, got %q", "LiteLLM", got)
+	}
+}
+
+func TestValidProviders_IncludesLiteLLM(t *testing.T) {
+	if !slices.Contains(ValidProviders(), "litellm") {
+		t.Fatalf("expected ValidProviders to include %q, got %v", "litellm", ValidProviders())
+	}
+}