NewAgent Observability — find failures before customers do.
Launching on Hacker News

Agent Observability at $10 per million spans.

OpenTelemetry native. Full prompt replay. Token tracking. No sampling.

Datadog charges $350/M spans. Langfuse charges per unit. Oodle charges $0.30/GB.

Get Started for FreeTry the Playground

Oodle

~$10

per million spans

Datadog

$350

per million spans

Langfuse

$60+

per million units

What you get

Your LLM call is just a span in a trace. Monitor it alongside your API, database, and cache calls.

Token Usage Tracking

Input, output, and reasoning tokens broken down per span.

Cost Estimation

Estimated cost per trace, per model, per service. 50+ models supported.

Prompt/Response Replay

Full multi-turn conversation replay: system, user, assistant, and tool messages.

Tool Call Analysis

See every tool invocation with schema, arguments, and response.

Quality Scoring

Automated quality scores flag low-confidence or off-topic responses.

LLM Span Highlighting

LLM spans stand out in your trace waterfall with provider icons and GenAI attributes.

See the math

Interactive calculator — plug in your volume and compare real monthly costs.

Ship in 15 minutes

Three steps from zero to full LLM visibility.

1

Instrument

Use any OpenTelemetry-compatible library — LangChain, Vercel AI SDK, OpenLLMetry, or roll your own.

2

Point OTLP at Oodle

One config line. Same endpoint you already use for backend traces. No new collector needed.

3

See everything

Tokens, cost, transcripts, quality scores — extracted automatically from your GenAI spans.

Trusted by

fast-growing startups, unicorns and public companies

Cureskin
HappyPathWisdom AIFuelLookoutZaggle
CureFit
Fello
Distacart
Workorb
Effective AI
different.ai
LabraBedrockdata
Cureskin
HappyPathWisdom AIFuelLookoutZaggle
CureFit
Fello
Distacart
Workorb
Effective AI
different.ai
LabraBedrockdata
Cureskin
HappyPathWisdom AIFuelLookoutZaggle
CureFit
Fello
Distacart
Workorb
Effective AI
different.ai
LabraBedrockdata

Frequently Asked Questions

What does '$10 per million spans' mean?

Oodle charges $0.30/GB for all traces — LLM or otherwise. A typical agent span is around 2KB, so one million spans is roughly 2GB, which costs about $0.60. Even with larger spans (tool calls, full prompt/response content), the cost stays well under $10 per million. Compare that to Datadog at $350 per million LLM spans.

How does Oodle price compare to Datadog Agent Observability?

Datadog charges $160/month base + $3.50 per 10,000 LLM spans. That's $350 per million spans. Oodle charges $0.30/GB with no base fee — roughly $10 per million spans. At 10M spans/month, Datadog costs ~$3,650 while Oodle costs ~$100.

Do I need a separate SDK?

No. Oodle consumes standard OpenTelemetry traces using the Generative AI semantic conventions. Point your OTLP exporter at Oodle — one config line, no new SDK.

Do you sample LLM traces?

No. At $10 per million spans, we don't need to. Retain 100% of your traces with full prompt and response content.

What LLM providers are supported?

Any provider your code calls — OpenAI, Anthropic, Google Gemini, Mistral, Cohere, Azure OpenAI, AWS Bedrock, and self-hosted models. Cost estimation ships with 50+ model pricing definitions.

Is there a free tier?

Yes. 50GB of traces per month, 100K active time series, and 50GB of logs — free forever. No credit card required.

Start sending traces in 15 minutes

Free tier included. No credit card required.

Get Started for FreeRead the Docs