Move off Humanloop recover and rebuild your Humanloop workflows.
EvalGuard has every Humanloop feature that matters — prompt editor, eval, deployments, human feedback, versioning — plus red-team security, LLM firewall, gateway, and FinOps in the same workspace. No vendor ownership by a model provider.
Why now, not later
Sunset timelines aren't yours to control
After an acquisition, the product roadmap — including any sunset date — is set by the acquirer, not by you. If Humanloop is in your production path, that's a hard-deadline risk worth planning for. Migrating on your own schedule is cheap; migrating during a forced cutover is an incident.
Your eval platform shouldn't be owned by a model vendor
Humanloop's platform was sunset in September 2025, after its team joined Anthropic. Whether a model vendor should own the platform you use to test its models for safety — or to compare them against OpenAI, Gemini, or open-source alternatives — is a question worth weighing. EvalGuard has zero model-vendor ownership.
Humanloop had prompts + evals. EvalGuard has six products.
Red team (300+ plugins), LLM firewall, gateway with 90+providers, OTel observability, FinOps cost tracking — all sharing one auth, one bill, one SLA. Consolidate vendors, don't just swap them.
Feature parity, at a glance
Everything you use in Humanloop has a direct equivalent — most are stronger in EvalGuard because the same platform also handles security, observability, and cost.
| Humanloop | EvalGuard | Result |
|---|---|---|
| Prompt Editor (side-by-side model comparison) | Playground + Prompt Optimizer across 90+ providers | Stronger |
| Datasets | Datasets (CSV/JSON import, versioned) | Parity |
| Evaluations (LLM-as-judge, human labels) | 200+ scorers incl. LLM-as-judge, pairwise, rubric-based | Stronger |
| Human Feedback / Annotation | Annotation Queues + Krippendorff's alpha | Parity |
| Deployments / Prompt Versioning | Prompt versions + A/B testing + shadow deploy | Stronger |
| Logs / Observability | OTel trace ingest + drift detection + cost attribution | Stronger |
| SOC 2 evidence / Enterprise SSO | SAML/OIDC SSO + SCIM + 50 compliance frameworks | Parity |
Migrate your project in three commands
Export your project from Humanloop (Humanloop owns the export — we never touch your Humanloop account), then convert it with the EvalGuard CLI. The importer reads your prompts, datasets, and evaluators into a runnable evalguard.config.json.
humanloop export project # or the "Export Project" download in the UInpx @evalguard/cli import:humanloop humanloop-export.jsonnpx @evalguard/cli eval:local evalguard.config.jsonThe importer maps every Humanloop evaluator it can (exact-match, contains, toxicity, bias, factuality, relevance, LLM-graded) to an EvalGuard scorer and flags any custom code evaluators so you can rewrite them. Prompt templates use the same {{var}} syntax — no rewrite needed.
eval:local runs entirely on your machine with your own provider key — no EvalGuard sign-up required. To run on the EvalGuard cloud instead (shared dashboards, run history, drift alerts), log in and use evalguard eval --project <id>.
Want a hand with the migration?
If your project has custom evaluators or a large dataset, send us your Humanloop export and we'll help you map it and validate the first run. Free.