Spanlens
Spanlens
Open-source LLM observability: log cost, latency, tokens, and agent traces for every OpenAI, Anthropic, Gemini, or Ollama call with a one-line baseURL swap. Self-hostable, MIT.
About Spanlens
Spanlens is an open-source LLM observability and monitoring platform. Developers point their provider baseURL at the Spanlens proxy, or use its drop-in SDK, and every model call is logged with cost, tokens, latency and full agent traces. The proxy speaks each provider natively, so streaming, tools and structured output keep working, and the site lists ten providers routed through the same endpoint, including OpenAI, Anthropic, Gemini, Groq, Mistral and OpenRouter.
The dashboard covers live cost per model and project, alerts on cost spikes, error bursts and latency drift, pinned prompt versions, exact-match caching via a header, and LLM-as-judge evals that can gate merges in CI. Agent traces show nested spans across tools, retrievals and model calls. PII detection, API key masking, OpenTelemetry ingest and export, and CSV, JSON or JSONL export are also included.
Spanlens is MIT licensed with no enterprise-only folder, so teams can self-host every feature with Docker on their own Postgres, or use hosted plans priced on request volume with seats included.
Best for
- Developers shipping LLM features to production
- Teams tracing multi-step AI agents
- Engineers tracking LLM spend across providers
- Organizations that want to self-host observability
Pros
- One-line baseURL change, no SDK rewrite
- Fully MIT licensed; self-hosting includes every feature
- Free plan needs no credit card
- Seats included in plans, no per-seat fee
Cons
- Free plan hard-blocks with 429 at 50K requests per month
- Free plan keeps logs for only 14 days
- Slack, webhooks and unlimited alerts require Team plan or higher
Pricing
Free plan with 50K requests/month, Pro at $29/month (100K requests, +$8 per 100K extra), Team at $149/month (1M requests, +$5 per 100K extra), custom Enterprise pricing, and free self-hosting.
Spanlens FAQ
- How do I instrument my app with Spanlens?
- Change your provider baseURL or use the drop-in SDK. Request logs, cost and latency are then captured, with streaming supported.
- Can Spanlens be self-hosted for free?
- Yes. It is fully MIT licensed, and running it with Docker gives every feature with no license fee, seat limit or usage cap.
- What counts as a request?
- One outbound LLM completion or embedding call counts as one request, including streamed responses and failed upstream calls. Tool runs and dashboard views do not count.
- Does it support OpenTelemetry?
- Yes. It supports OTLP/HTTP ingest and export, and adds LLM-specific attributes to LLM spans.
Compare Alternatives
See what other makers use instead of Spanlens
99finder.io
Every sales opportunity starts with a name—but without verified contact data, it goes nowhere. 99finder.io bridges that gap. With AI-powered discovery, it transforms your name lists into verified websites and contacts that drive real conversions.
AISDR
Outreach that feels human. AiSDR crafts timely, personalized messages at scale.