Service Description
Last updated: September 2026
What traceroot is
traceroot is the open-source platform that continuously improves your AI agents. You instrument your agent with our SDK or send OpenTelemetry data, and traceroot records each run as a trace: the model calls, tool calls, inputs, outputs, timing, and cost. On top of those traces it provides:
- Sessions and traces: browse, search, and filter runs; view the full call tree of any run.
- Detectors: rules that watch incoming traces for conditions you define and raise findings, with optional alerts to Slack.
- Root-cause analysis: an agent that reads a finding, the trace, and optionally your connected repository, and proposes a diagnosis.
- Evaluations: datasets, test cases, and scored runs to measure an agent’s behaviour over time.
- Dashboards: widgets over your trace data.
- An open-source core: the platform is developed at github.com/traceroot-ai/traceroot and can be self-hosted with our Helm chart and Terraform module.
How you connect
| Client | Use |
|---|---|
| TypeScript SDK | Instrument Node.js agents; auto-instrumentation for OpenAI, Anthropic, LangChain, Bedrock, Vercel AI SDK, Mastra |
| Python SDK | Instrument Python agents |
| CLI | Query traces and sessions from the terminal or CI |
| Claude Code and Codex plugins | Trace coding-agent sessions |
| OpenTelemetry | Send spans from any OTel-instrumented service |
| Public API | Read traces, sessions, detectors, and evaluations with a project API key |
Plans
| Plan | Query window | Included |
|---|---|---|
| Free | 15 days | 2 seats, 50k events per month, 30 chat runs, 30 RCA runs, 100 detector runs |
| Starter | 30 days | Unlimited seats, 150k events per month with metered overage, 100 chat and 100 RCA runs, unlimited detector runs |
| Pro | 90 days | Everything in Starter, higher rate limits (20k ingest and 1k dashboard requests per minute) |
| Enterprise | Custom | Custom limits, contract terms, and compliance documentation |
The query window limits how far back the product shows data; it does not delete data. Current prices are shown at checkout.
Pricing and overage details
Starter includes 150k events, 100 chat runs and 100 RCA runs per month.
- Additional events: $4 per 50k events.
- Additional chat runs: $10 per 100 runs.
- Additional RCA runs: $10 per 100 runs.
- Hosted LLM: token cost × 1.05 on overage. Bring your own keys is supported on all plans.
Pro inherits Starter features with 90-day retention and higher rate limits. Enterprise includes custom retention, SLA support and Discord + Slack + SLA.
What traceroot is not
- Not a model provider: agent and analysis features call Anthropic or OpenAI with your key or ours, and their availability depends on those providers.
- Not a general log or metrics platform: it stores agent traces, not application logs or infrastructure metrics.
- Not a system of record for your source code: repository contents are read per run and not retained.
Data location and retention
Hosted in AWS us-east-1. Customer data is kept while the workspace is active and deleted on request within 30 days; see the Privacy Policy.
Support and communication
- Support: [email protected] and the community Discord. Enterprise plans include a dedicated channel.
- Product changes are announced in release notes on GitHub; breaking changes to SDKs or the API are announced at least 30 days in advance and versioned.
- Incidents affecting the hosted platform are communicated by email to workspace owners.