LLM observability

See every LLM call your app makes.

One line of code traces requests, tokens, cost, and latency across 50+ AI libraries — OpenAI, Anthropic, LangChain, vector databases, and more. Your data lands in an isolated project only you can see.

Free during beta · up to 3 projects per account · no card needed

quickstart.py

# pip install tidal-telemetry
import tidal_telemetry as tidal

tidal.init()  # reads TIDAL_API_KEY + TIDAL_PROJECT

# done - every LLM call below is now traced:
# tokens, cost, latency, prompts, errors
client.chat.completions.create(...)

Live in three steps

Sign up

Create an account and add up to three projects. Each one gets its own endpoint and ingest key.

Add one line

Install tidal-telemetry for Python and call tidal.init(), or connect any OpenTelemetry exporter to your project endpoint. The Node.js wrapper is still in preparation.

Watch every call

Requests, token counts, cost estimates, latency, and full traces appear in your dashboard as they happen — across every model and library you use.

Everything you need to run LLMs in production

Built on OpenTelemetry and the Apache-licensed OpenLIT SDK — open standards end to end, no lock-in.

Auto-instrumentation

50+ AI libraries traced out of the box: OpenAI, Anthropic, Mistral, Bedrock, LangChain, LlamaIndex, vector DBs, and agent frameworks.

Token & cost tracking

Every span carries input/output token counts and a cost estimate from per-model pricing tables — see spend per app, per environment, per day.

Isolated per-project storage

Each project gets its own database and its own ingest key. Your traces are never mixed with anyone else's — enforced at the storage layer.

Standard OpenTelemetry

Ingest is plain OTLP over HTTPS. The SDK is optional — point any OpenTelemetry exporter at your project endpoint and it just works.

Ask your agent

Connect the hosted MCP server to check account status, verify ingestion, send a test trace, read usage, and troubleshoot. See the agent setup guide.

Protected ingest

Bearer-token auth per project and per-tenant rate limiting keep your pipeline healthy and your keys revocable at any time.

Pricing

Simple while we're in beta: everything is free, up to 3 projects per account. Beta users keep a free plan.

Live now

Beta

$0 while in beta

All features included
Python SDK + raw OTLP from any language
Isolated project database
600 requests/min fair-use ingest
MCP usage reports

Planned

Pro

$49 / month

Everything in Beta
Higher rate limits
Extended trace retention
Multiple projects & teammates
Email support

Get notified

Planned

Scale

Custom

Everything in Pro
Custom limits & retention
Managed cost optimization by LLM CFO
Priority support

Talk to us

Planned tiers and prices may change before launch. Cost figures shown in dashboards are estimates from per-model pricing tables, not invoices.

Frequently asked questions

Do I have to use your SDK?

No. Ingest is standard OTLP over HTTPS. The tidal-telemetry SDK is a thin convenience wrapper around the open-source OpenLIT SDK — you can point any OpenTelemetry exporter at your project endpoint with a bearer token and it works the same.

Is my data isolated from other customers?

Yes. Every project gets its own database and its own ingest key. A key for one project is rejected on any other project's endpoint, and dashboards only ever query your own project's database.

Which languages are supported?

Python (pip install tidal-telemetry) has one-line setup. Node.js can export through OpenTelemetry today; its Tidal wrapper is not published yet. Any language that speaks OpenTelemetry can export to your project endpoint.

What does tracing capture?

Model, prompt/completion metadata, input and output token counts, an estimated cost per call, latency, and errors — streamed as OpenTelemetry spans you can slice by application, environment, and time.

How is this related to LLM CFO?

Tidal Telemetry is LLM CFO's observability product. Telemetry shows you what your LLMs spend; LLM CFO's managed service then cuts that spend with routing, caching, and model-mix changes — priced on verified savings.