Traceloop maintains OpenLLMetry, the open-source OpenTelemetry standard for LLM instrumentation, plus a hosted monitoring platform on top of it. It's built to get traces out of your stack via standard OTel spans — evaluation-based guardrails and CI checks now exist, but neither re-executes a captured step or produces a signed audit record.
| Capability | Traceloop | Runback | Note |
|---|---|---|---|
| Broad auto-instrumentation via OpenTelemetry | ✓ | partial | Runback ingests standard GenAI OTel spans too |
| Read and search traces | ✓ | ✓ | |
| Open-source instrumentation layer | ✓ | partial | Runback's OTel ingestion is open; the replay/audit core is not |
| Re-execute any step from captured inputs | ✗ | ✓ | The OTel GenAI spec standardises what gets emitted, not what happens to it once stored — nothing on the receiving end re-runs a step from it |
| Live policy enforcement — block before it runs | partial | ✓ | Guardrails evaluators can block LLM output in real time — narrower than a tool-call-level policy engine, and evaluation-based rather than a simulated-against-history gate. Runback decides each tool call server-side, from its TypeScript and Python SDKs (not yet Go or an MCP proxy), against the policy version assigned to the agent; each assignment fails closed or open during an outage (closed by default for agents declared critical, once the SDK has fetched its policy bundle), and decisions are recorded and sealed — those the SDK makes on its own once Runback answers again |
| CI release gate | partial | ✓ | A GitHub App runs evals on every PR and posts results — an evaluation-based check, not Runback's simulate-against-production-history gate |
| Signed, tamper-evident audit export | ✗ | ✓ | Spans reach the backend over an open wire format, but nothing seals them afterward — the standard defines transport, not chain-of-custody once stored |
| Regulatory mappings (EU AI Act · CPS 230 · NIST) | ✗ | ✓ |
It's built to get traces out of your stack via standard OTel spans — evaluation-based guardrails and CI checks now exist, but neither re-executes a captured step or produces a signed audit record. Most teams use both — Traceloop for what it's already good at, Runback for deterministic incident replay and the signed record.
Source: Traceloop's public documentation, https://www.traceloop.com/docs, read 15 September 2026. Products change: report an error to legal@runback.dev and we will correct it. Names and trade marks belong to their owners; Runback is not affiliated with, endorsed or sponsored by any of them.
No signup. Walk the scripted failing demo run step by step and see what the model was given.