Alpha Release - Completely Free
Know How Your AI Agent Works
LLM SPY records traffic between your AI Agent and LLMs, revealing exactly what it does - every prompt, tool call, and decision.
llm-spy-monitor
ANTHROPIC_BASE_URL="https://llms.klmn.sh" \ ANTHROPIC_CUSTOM_HEADERS="x-proxy-auth: KEY" \ claude > Build me a LangChain agent that teaches me how to cook sushi
See the results
https://llms.klmn.sh/workspace-1/requests
How It Works
Set up in minutes, monitor everything instantly
Create a Proxy
Point your AI agent to LLM SPY instead of the LLM provider directly. Just change one environment variable and you're monitoring.
Monitor Everything
View all LLM traffic in real-time. See prompts, responses, tool calls, token usage, and execution flow.
Share Conversations
Share specific LLM conversations with your team via secure, secret links. Perfect for debugging, collaboration, or demonstrating AI behavior.
Example Use Case
Claude Code Monitoring
Monitor what Claude Code actually does by setting ANTHROPIC_BASE_URL
See all prompts and context
Track tool usage patterns
Monitor token consumption
Examples
Select an upstream, key, and platform to view connection instructions.
Select Upstream
anthropic
Select API Key
sk_demo_api_key
Select Platform
Select a platform
Why LLM SPY?
Complete transparency and control over your AI agents
Complete Visibility
Know exactly what your LLM does - every prompt, response, tool call, and decision. No more black box AI.
- Full prompt history
- Tool execution logs
- Token usage tracking
Instant Setup
Just change one environment variable and you're monitoring. No SDK integration, no code changes required.
- One-line setup
- Works with any agent
- Zero performance impact
LLM Translation Gateway
Coming SoonUse any LLM through any interface. Your app talks to 'Anthropic' but actually uses GPT-4.
- Format conversion
- Provider switching
Coming Soon
Session Intelligence
Automatically groups related LLM interactions into coherent sessions, giving you the full conversation context.
Smart Conversation Tracking
LLM SPY intelligently detects and groups related requests into sessions. See the complete flow of multi-turn conversations, not just isolated API calls.
Automatic Session Detection
Identifies conversation patterns and links related messages together. No manual tagging or session IDs required.
Full Conversation Context
View entire conversation threads with preserved context. Understand how your agent builds on previous exchanges.
Session #A7B2
Request 1 • 14:23:01
→ "Create a Python function to parse JSON"
Response 1 • 14:23:02
← "def parse_json(data): ..."
Request 2 • 14:23:15
→ "Add error handling to that function"
Response 2 • 14:23:16
← "try: ... except JSONDecodeError: ..."
Automatically grouped into one session
Coming Soon
LLM Translation Gateway
Make any LLM speak any language. Your app talks to "Anthropic" but actually uses GPT-4, or vice versa.
Universal Format Translation
LLM SPY acts as a translation layer between your application and any LLM provider. Switch between models without changing a single line of code.
Automatic Format Conversion
Seamlessly converts between Anthropic, OpenAI, Google, and other LLM formats. Use any model with any interface.
Full Request Visibility
See exactly how requests are translated between formats. Debug and understand the conversion process in real-time.
# Your app thinks it's using Claude...
ANTHROPIC_BASE_URL="https://llms.klmn.sh" \
ANTHROPIC_CUSTOM_HEADERS="x-proxy-auth: key" \
claude
> Build me a LangChain agent that teaches me how to cook sushi
# But actually uses GPT-4 via translation gateway
Frequently Asked Questions
Everything you need to know about LLM SPY
Pricing
Free to Use
$0
No cost while hosting is manageable
It's free while it doesn't incur significant hosting or storage costs. We may start charging if it becomes an issue, but will keep it affordable.
Start NowReady to See What Your AI Really Does?
Join the alpha and get complete visibility into your AI agents
Get Started for Free