Domain Rating
by Ahrefs
Domain Rating by Ahrefs45
Description
Spanlens is an observability platform for LLM applications that works with OpenAI, Anthropic and Gemini, and with other providers such as Groq, Mistral, xAI and OpenRouter. To start, you change one line of config, the provider base URL, and your SDK call stays the same. Every model call is logged with its cost, latency and tokens, and you can see spend per model and project, including cache savings. It flags cost spikes, error bursts and latency drift, and shows the model and project responsible. You can pin prompt versions and compare cost and quality on real traffic. Exact-match caching can be switched on with a header, so repeat calls skip the provider. Evals can run in CI to gate a merge, and agent traces show nested spans across tools, retrievals and model calls. PII detectors flag things such as card numbers and email addresses without blocking the request. Spanlens is open source under the MIT licence and can be self-hosted with Docker, or used as a hosted service with a free plan and a Pro plan.…
Preview
Pricing
- Free$0/mo
- Pro$29/mo
Highlights
- One-line base URL change to start logging
- Cost, latency and token logs per model
- Anomaly alerts and PII detection
- Agent traces and evals in CI
- Pinned prompt versions and exact-match caching
- MIT licensed and self-hostable with Docker
Ratings & Reviews
No ratings yet
Be the first to rate Spanlens.
People also visit
Spanlens FAQ
How does instrumentation work?
Change your provider base URL or use the drop-in SDK to capture request logs, cost and latency. Streaming responses are supported, and request logs are saved asynchronously.
What about latency overhead?
Request logs are saved asynchronously, so your request does not wait for log storage. The site links to benchmarks that show the measured CPU work and how it was measured.
How do you handle PII?
PII detectors for things like SSNs, card numbers, emails, IBANs and passports run at log time and flag matches for review without blocking the request. API keys that appear in prompts are masked before they are stored. You can opt out of storing prompt bodies for a call with a header.
Do you support OpenTelemetry?
Yes. It supports OTLP/HTTP ingest and export, so existing OpenTelemetry tracing flows into the same span store.
What are the best Spanlens alternatives?
Popular alternatives to Spanlens include GoModel, Langfuse and Inferly. See all Spanlens alternatives



