Every tool, token, session - tracked

Claude Code and Codex, one OpenTelemetry config change, full visibility in minutes.

AI Agent Observability dashboard showing Claude Code session costs, token usage, and tool call analytics

Your engineers use AI agents daily. Do you know the cost?

AI coding agents generate real spend, real risk, and real waste. Oodle surfaces all three without any instrumentation work.

Cost Visibility

Track spend by user, model, team, repository, and session. Identify top spenders and wasteful patterns before the bill arrives.

Session Intelligence

AI-powered insights on every session: context runaway, cache efficiency, model mismatch, vague prompts. Actionable, not just data.

Zero Instrumentation Effort

Native OpenTelemetry from Claude Code and Codex. One env var or config file, no SDK, no code changes.

Cost. Sessions. Tools. Covered.

Full visibility into how your engineering team uses AI coding agents.

Cost Breakdown by Dimension

Slice spend by user, model, query source, agent, skill, team, and git repository. Know who spends what and why.

Session-Level AI Insights

Every session gets automated analysis: context growth warnings, cache efficiency scores, model mismatch detection, and tool-heavy session flags.

Tool Usage Analytics

Track every tool call: frequency, duration, result size. Spot MCP tool bloat, bash anti-patterns, and context pollution.

Cache Efficiency Tracking

Monitor cache hit rates across users and models. Identify ghost cost from sessions that repeatedly re-prime cache after idle gaps.

Cost-per-PR Leaderboards

Tie AI spend to engineering output. Compare cost per pull request, sessions with commits vs without, and cost per line of code.

Turn-by-Turn Event Replay

Expand any session into a searchable turn-by-turn log: prompts, API calls, tool decisions, and tool results with raw JSON.

Trace Correlation

Link Claude Code OTEL traces directly to session context. Debug slow tools or failed API calls with full distributed trace waterfall.

How It Works

Three steps from zero to full AI agent visibility.

Enable Telemetry

Add OTEL env vars to ~/.claude/settings.json or ~/.codex/config.toml. Oodle generates the config for you from the integration page.

Sessions Flow In

Token usage, costs, tool calls, and prompts stream to Oodle via OTLP. No collector, no pipeline to maintain.

Insights Surface Automatically

Pre-built dashboards, AI analysis, and cost alerts. Every session analyzed for waste, risk, and optimization opportunities.

See It With Your Team's Sessions

Book a call to see how Oodle tracks your AI agent usage, cost, and efficiency.

Schedule Demo View Docs

Frequently Asked Questions

What agents are supported?

Claude Code and OpenAI Codex today. Both are GA. Oodle automatically detects the agent type from the telemetry and routes sessions to the correct dashboards and analysis views.

How long does setup take?

Under 5 minutes. For Claude Code, add OTEL env vars to ~/.claude/settings.json. For Codex, add an [otel] block to ~/.codex/config.toml. No SDK, no code changes, no collector to maintain.

What data is collected?

Metrics (tokens, cost, sessions, tool calls, lines of code, commits, pull requests), structured log events (prompts, API calls, tool results), and traces. Prompt and tool detail logging are opt-in.

Do I need to install a collector?

No. Both Claude Code and Codex export OpenTelemetry data directly to Oodle's OTLP endpoint. No collector, no pipeline, no infrastructure to maintain.

Can I control what gets sent?

Yes. Prompt logging is opt-in via OTEL_LOG_USER_PROMPTS. Tool detail logging is opt-in via OTEL_LOG_TOOL_DETAILS. You can also scope telemetry to specific teams and repositories using OTEL_RESOURCE_ATTRIBUTES.

How does this relate to LLM Observability?

LLM Observability tracks model calls inside your application traces (APIs calling OpenAI, Anthropic, etc). AI Agent Observability tracks the coding agents your engineers use day-to-day (Claude Code, Codex). Different audiences, complementary products. Learn more about LLM Observability.