Changelog
What's new in LLMWatch
June 20, 2026
Local LLM support
- •Route requests from LM Studio, Ollama, or any OpenAI-compatible local server through the proxy
- •Pass x-llmwatch-local-url header with your local server URL to get cost and latency tracking for local models
June 21, 2026
GraphQL API
- •New GraphQL endpoint at /api/graphql for flexible, query-based access to logs and stats
- •Available as a code example in the View Code modal
June 20, 2026
Caching opportunity detection
- •Automatically detects when you send the same prompt multiple times
- •Shows potential cost savings if you implement caching on repeated requests
June 20, 2026
A/B testing support
- •Tag requests with x-llmwatch-experiment header to compare prompt or model variants
- •See cost and latency side by side for each experiment tag on the logs dashboard
June 20, 2026
Anomaly detection
- •Get an automatic warning when today's spending is more than 2x your recent daily average
June 19, 2026
One-click Slack integration + public badges
- •Connect Slack with one click — no more copying webhook URLs manually
- •Get a public LLMWatch badge to embed in your README or website
June 19, 2026
Custom domains, code snippets, cost projections
- •Pro users can now use a custom domain for the proxy endpoint
- •One-click integration code in Node.js, Python, or cURL
- •Projected monthly cost based on current usage trends
June 19, 2026
Error rate webhooks + public API
- •Set an error rate threshold to get webhook alerts when reliability drops
- •New public API endpoints (/api/v1/logs, /api/v1/stats) for programmatic access
- •Search logs by prompt, response, or model name
- •Model comparison table — see cost and latency side by side across models
June 19, 2026
Error rate tracking
- •Failed and timed-out requests are now logged alongside successful ones
- •New Error Rate metric on the logs dashboard to spot reliability issues early
June 17, 2026
Team collaboration
- •Invite teammates to view your project (read-only access)
- •Project owners can manage and remove team members
- •Invited members can leave a shared project anytime
June 17, 2026
Webhook notifications + CSV export
- •Get real-time alerts via Slack, Discord, or any webhook URL when a request exceeds your cost threshold
- •Export your full request history as CSV
June 16, 2026
Streaming support
- •The proxy now supports streaming responses (stream: true)
- •Tokens and cost are calculated after the stream completes
- •No changes needed in your code — streaming just works
June 16, 2026
Multi-provider support
- •Added support for Anthropic, Gemini, Mistral, Groq, DeepSeek, Together AI, and OpenRouter
- •Automatic provider detection based on model name
June 15, 2026
Security improvements
- •Added rate limiting (60 requests/minute per API key)
- •Encrypted prompt and response data at rest
- •Added API key rotation
- •Webhook signature verification for billing events
June 13, 2026
Request detail view + cost charts
- •Click any log to see full prompt, response, and latency breakdown
- •Daily cost chart on the logs page
- •Separated proxy overhead from actual API latency
June 11, 2026
Cost alerts
- •Set a monthly budget per project
- •Get an email when you hit 80% of your limit
June 10, 2026
Launch
- •LLMWatch is live
- •Track OpenAI API costs by changing one line of code
- •Free tier: 1,000 requests/month