Oodle.ai: Columnar storage engine for LLM agent observability and tracing

What it does

Oodle.ai is a columnar storage engine designed for observability of large language model agents. It ingests and stores agent traces, logs, and metrics without sampling, meaning every session is preserved in full. The system stores data in S3 using a custom columnar file format similar to Parquet, then runs queries on serverless compute. The core offering is agent observability: identifying which agent interactions failed, running evaluations, and tracing tool calls across application and infrastructure layers.

Who it is for

The product targets teams building production AI agents who need to debug and improve agent behavior at scale. The homepage shows customers processing 3M+ agent traces daily. It integrates with major LLM frameworks including OpenAI, Claude, LangChain, Vercel AI SDK, Pydantic AI, and others. Setup is described as two commands pasted into a coding agent, or pointing OpenTelemetry at Oodle.

Pricing

The site does not show prices.

How it stands out

Oodle's core differentiator is storing 100% of agent traces without sampling while keeping costs low. This is achieved through two mechanisms: storing data in object storage (claimed to be 20× cheaper than disk) and compression in the custom columnar format (600× for metrics, 20× for logs and traces). Queries run on serverless compute that scales up per query and shuts down after completion, avoiding fixed provisioned capacity costs. The homepage claims this approach enables both fast queries and inexpensive storage at scale.

The product offers three deployment options: managed SaaS, bring-your-own-bucket (S3), and bring-your-own-cloud (VPC). It supports industry compliance standards including SOC 2 Type II, ISO 27001, and GDPR.

What a founder should check

A founder building a competitor should verify: (1) whether the claimed compression ratios and query performance hold under real customer workloads and at different data volumes, as these are the core value propositions; (2) how sticky the integration is given that Oodle works as a drop-in replacement for OpenTelemetry and Langfuse endpoints—what prevents customers from switching once integrated; (3) pricing and unit economics compared to incumbent observability platforms like Datadog or New Relic, especially whether the serverless model remains cost-advantageous as query concurrency grows or data retention extends beyond what competitors offer.

Thinking of building something like this?

Every launch here is a competitor to somebody's idea. If yours is close, check it against the market before you build: the Full Check names the rivals, the prices and the gaps.