| What it does | Open-source agent harness engineering platform built on OpenTelemetry
| AutoML that just works. Keep training local.
|
|---|
| Stage | Startup
| MVP
|
|---|
| Founded | 2024
| Not given
|
|---|
| Based in | IN
| Not given
|
|---|
| Pricing model | Open source
| Free trial
|
|---|
| Pricing | Open source (Apache 2.0): Free self-hosted with unlimited usage. Enterprise Edition: Contact sales for pricing. Cloud: Coming soon.
| Free beta with no billing. Paid plans launching in the future; pricing not yet announced. Join the waitlist for paid plans without payment or purchase commitment.
|
|---|
| Key features | - OpenTelemetry-native LLM tracing for every call, tool, and retrieval
- Agent evals with LLM-as-a-judge and heuristics
- Prompt Hub for versioning and deployment without code changes
- Guardrails for prompt injection, jailbreak, and sensitive topic detection
- Vault for encrypted API key management
- OpenGround for side-by-side LLM model comparison
- GPU monitoring for utilization, memory, temperature, and power
- Coding agent observability for Claude Code, Codex, and Cursor
- Cost tracking and latency monitoring
- 56+ integrations with LLMs, frameworks, and vector databases
| - Python SDK for tuning suggestions
- Local model training with remote suggestions
- Supports classification and regression
- Included CPU demo workloads
- Efficient observation budget allocation
- Model ranking and selection
- Reference scheduler implementations
- HTTPS secure communication
|
|---|
| What makes it different | - Self-host free forever under Apache 2.0 with unlimited usage and no per-trace fees
- OpenTelemetry-native traces exportable to Grafana, Datadog, or any OTLP backend
- Zero-code instrumentation with eBPF controller, native SDKs, or GPU collector
- Unified platform for traces, evals, guardrails, and prompts on the same production data
- Full data ownership and no vendor lock-in with exportable OpenTelemetry traces
| - Trains locally, not on the service—your models and data stay on your computer
- Efficient tuning using only half the observation budget of full training
- Outperforms reference schedulers in benchmarks on diverse datasets
- Free beta with no billing or account costs
|
|---|
| Community votes | 0 | 0 |
|---|
| DR (Domain Rating by Ahrefs) | 65 ● no change since last check | Not checked yet ● no change since last check |
|---|
| Trust Flow (Majestic) | 23 | 3 |
|---|
| Citation Flow (Majestic) | 39 | 13 |
|---|
| Referring domains (Majestic) | 483 | 2 |
|---|
OpenLIT
hyperopt.dev