Helicone has been in maintenance mode since March 2026 and is not taking new cloud sign-ups. For a new deployment, Datadog is usually the practical choice of the two; for existing Helicone users, the useful question is what to migrate to and how, including what replaces Helicone's gateway.[1][6]
Datadog Agent Observability, formerly LLM Observability, is an actively developed hosted module of the Datadog platform, while Helicone is an open-source LLM request logger and gateway that has been in maintenance mode since March 2026, with new cloud sign-ups disabled. Mintlify acquired Helicone in March 2026; on 19 October 2026 Helicone said it is no longer part of Mintlify, with no new owner named. For a new deployment Datadog is usually the practical choice of the two; for existing Helicone users the question is how to migrate.
| Decision factor | Datadog | Helicone |
|---|---|---|
| Status | Actively developed; renamed from LLM Observability in June 2026 | Maintenance mode since March 2026; new cloud sign-ups disabled |
| Ownership | Datadog, a public company (Nasdaq: DDOG) | Unclear: no longer part of Mintlify as of 19 October 2026, no new owner named |
| How traces arrive | SDKs, auto-instrumentation, HTTP API or OTLP | Proxy or AI Gateway in the request path, or async OpenLLMetry-based SDKs |
| Agent tracing | Spans for tools, handoffs and multi-agent runs | Header-based session grouping, not a span model |
| Evaluation | Experiments, LLM-as-a-judge on production spans, Annotation Queues | Not an evaluation framework, by its own docs; scores posted from outside |
| Gateway | None; documents tracing a gateway you run | AI Gateway: one endpoint for 100+ providers with failover and caching |
| Self-hosting | None documented | Yes: Apache-2.0, Docker or Kubernetes |
| Free tier | Free: 40K LLM spans a month, 15-day retention | Hobby: 10K requests, 7-day retention, but sign-ups are disabled |
| Capability | Datadog | Helicone | Prefactor |
|---|---|---|---|
| Observe | |||
| LLM call tracing | ✓ | ✓ | ✓ |
| Multi-step agent tracing (tool calls, nested spans) | ✓ | Partial | ✓ |
| Cost tracking | ✓ | ✓ | ✓ |
| OpenTelemetry support | ✓ | Partial | ✓ |
| Evaluate | |||
| Offline evals (datasets, experiments) | ✓ | Partial | ✓ |
| Online evals on production traffic | ✓ | Partial | ✓ |
| LLM-as-a-judge scoring | ✓ | Partial | ✓ |
| Code-based evaluators | ✓ | — | ✓ |
| Human review and annotation | ✓ | Partial | ✓ |
| Prompt versioning and playground | ✓ | ✓ | — |
| Act on findings | |||
| Runtime controls on live agent actions | Partial | Partial | ✓ |
| Gateway or proxy in the request path | — | ✓ | — |
| Licence and deployment | |||
| Open source | Partial | ✓ | Partial |
| Self-hosting | — | ✓ | Partial |
Mintlify announced its acquisition of Helicone on 3 March 2026 and said the product would move to maintenance only: patches, bug fixes and new model support. The cloud sign-up page has shown that sign-ups are disabled since the same day, and the changelog shows no feature entries after November 2025. On 19 October 2026 Helicone's site began stating that it is no longer part of Mintlify, with details to follow and no new owner named.
Commits continue: in October 2026 Helicone turned off Python evaluators, rate-limited LLM-as-a-judge calls and hardened its self-hosted stack. In May 2026 an AWS account lock took its dashboard and log processing down for about four days while the proxy stayed up. None of that suggests a product to start on today. Datadog Agent Observability, by contrast, shipped Patterns and Bits Evals in Preview in June 2026 and added a local trace viewer and OpenTelemetry ingestion without its SDK by October.
Start with how you send data today. Helicone's async logging SDKs are built on OpenLLMetry instrumentation, and Datadog's OTLP intake recognises OpenLLMetry, so that part of a migration may be a change of export target rather than new instrumentation. Test it on one service first; Datadog notes a 3 to 5 minute ingestion delay for OTLP traces.
If you use Helicone as a proxy or through its AI Gateway, logging is only half of what it does: it also routes to 100+ providers with failover, caching, rate limits and key handling. Datadog has no gateway of its own. It documents how to trace one you run and auto-instruments LiteLLM, so you will need a separate gateway if you rely on those features.
Prompts move too. Helicone versions prompts with environments and deploys them by ID through its gateway; Datadog has a versioned prompt registry retrieved at run time from Python, Go or JavaScript, with per-environment deployment. Plan to export and re-create prompts rather than expect an import path, since neither sheet documents one. Teams that cannot move yet can self-host the Apache-2.0 codebase, accepting maintenance-only updates.
You gain evaluation. Helicone's own docs say it does not run evaluations for you; scores from tools such as Ragas are posted to requests from outside, and Python evaluators were turned off in October 2026. Datadog runs experiments on versioned datasets, LLM-as-a-judge evaluators on sampled production spans and Annotation Queues for human review. You also gain an agent span model: Datadog shows tools and handoffs for OpenAI Agents SDK, LangGraph and CrewAI, where Helicone groups requests into sessions with path headers.
You lose self-hosting and the request-path gateway. Datadog is hosted only, needs a Datadog account and is not available on its US government sites. Helicone's prompt injection blocking, which works with OpenAI models only, has no equivalent inside Datadog's module; Datadog sells blocking separately as AI Guard, in Limited Availability.
Cost tracking carries over in both directions. Helicone prices requests for 300+ models with cost alerts; Datadog calculates cost from token counts for over 800 text-based models and accepts custom values for self-hosted ones.
As of 20 October 2026, Helicone's pricing page lists Hobby free with 10K requests and 7-day retention, Pro at $79 a month and Team at $799 a month, but new cloud accounts cannot be opened on any plan. Datadog's Free plan covers 40K LLM spans a month with 15-day retention, and Pro is $160 a month billed annually for the first 100K LLM spans, then $3.50 per 10K, with only spans that call an LLM provider billed. Helicone counts requests and storage, while Datadog counts LLM spans, so measure your own volumes in both units. Check Datadog's pricing and Helicone's pricing for current terms.
We sell the layer this section describes. Read it with that in mind.
Stated plainly, since Prefactor publishes this page: Prefactor is an agent evaluation platform that scores each step of a run and can act on what it finds through runtime policies on tool calls and data access, approval holds and a kill switch scoped to a run, an agent or a team. Helicone's blocking covers prompt injection on OpenAI models only, and Datadog's sits in AI Guard, a separate product; neither holds an agent action for approval. Prefactor has no prompt playground or prompt registry, no gateway and is hosted, so a team migrating Helicone's gateway or prompts will need something else for those. See the full comparison of 15 platforms.
Reviewed against public product and documentation pages on October 20, 2026. If a vendor has changed a feature, product name, or positioning since then, send a correction and we will update it. Numbered source links in the page body point to the ordered sources below.
Prefactor scores each step of every agent run and can hold or stop an action when a score says it should not happen. Start on the free tier: 5,000 spans a month, no credit card required.
Prefactor helps teams observe, evaluate, and improve their AI agents in production, across every framework and provider.