Both are Apache-2.0 and can be self-hosted. Helicone is a request logger and AI gateway in maintenance mode, closed to new cloud customers; Opik is an actively developed platform for agent tracing, evals and prompt work. For anything new, that status difference settles it.[3][4]
The main difference between Helicone and Opik is status: Helicone has been in maintenance mode since Mintlify acquired it in March 2026, with new cloud sign-ups disabled, and on 19 October 2026 it said it is no longer part of Mintlify, with no new owner named. Opik, from Comet, is an actively developed Apache-2.0 tracing and evaluation platform. For a new deployment Opik is usually the practical choice; existing Helicone users should plan a migration.
| Decision factor | Helicone | Opik |
|---|---|---|
| Status | Maintenance mode since March 2026; cloud sign-ups disabled | Active; Opik 2.0 shipped June 2026, with changelog releases through October |
| Ownership | Unclear: said on 19 October 2026 it is no longer part of Mintlify; no new owner named | Comet ML, independent; reports $70M in total funding |
| Core job | Request logging and an AI gateway for 100+ providers | Tracing, evaluation and prompt versioning |
| Evaluation | Not an evaluation framework, per its docs; Python evaluators turned off October 2026 | Offline experiments and online rules with 30+ built-in metrics |
| Agent tracing | Partial: sessions grouped by request headers | Each step traced, including tool calls, with thread grouping |
| Gateway | Yes, with failover, caching and rate limits | Beta, for development and testing only |
| Self-hosting | Yes, Apache-2.0; gateway deployed alongside | Yes, Apache-2.0; Kubernetes for production |
| Capability | Helicone | Opik | 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 | ✓ | Partial | — |
| Licence and deployment | |||
| Open source | ✓ | ✓ | Partial |
| Self-hosting | ✓ | ✓ | Partial |
Mintlify announced its acquisition of Helicone on 3 March 2026 and said the product would move to a maintenance-only phase covering patches, bug fixes and new model support, and that it would help customers migrate to other platforms. Cloud sign-ups have shown 'Sign ups are disabled' since the same day, and the public changelog has no feature entries after November 2025.
On 19 October 2026 Helicone replaced its acquisition messaging with a banner saying it is no longer part of Mintlify, with details to come. No new owner or structure has been named. The same week, Python evaluators were turned off and judge calls were rate limited. Earlier, in May 2026, an AWS account lock took the dashboard and log processing down for about four days while the proxy stayed up. For a new deployment, no cloud account can be opened on any plan, and self-hosting means depending on a codebase whose future owner is not yet known.
Helicone's core is its AI Gateway: one OpenAI-compatible endpoint for more than 100 providers, with automatic failover, caching, custom rate limits and bring-your-own keys. It logs every request, tracks cost across more than 300 models with cost alerts, and versions prompts with environments and deployment by prompt ID. Agent runs are grouped into sessions through request headers, a path-based hierarchy rather than a span model. Its gateway can block prompt injection and jailbreak attempts, for OpenAI models only. Its documentation says plainly that Helicone is not an evaluation framework.
Opik is built for evaluation. It traces each agent step, including retrieval and tool calls, groups conversations into threads, and scores sampled production traffic with online rules using more than 30 built-in metrics, custom judges or custom Python metrics. It adds annotation queues, a Prompt Library with environment-tagged versions, a Prompt Optimizer and guardrails your code calls. Its own LLM Gateway is in beta and documented for development and testing only.
The products overlap least on the gateway, so start there. If Helicone sits in your request path for routing, failover or caching, do not plan to replace it with Opik's beta gateway. Opik integrates with third-party gateways, including LiteLLM, Portkey, OpenRouter, Kong AI Gateway, TrueFoundry and Vercel AI Gateway, and lists Helicone itself among them, so you can keep traffic flowing through Helicone while tracing moves to Opik, then replace the gateway as a separate step.
Instrumentation moves from Helicone's headers or async SDKs to Opik's SDKs for Python, TypeScript, Java or .NET, or to OTLP over HTTP. Neither vendor documents an import path for prompts, so prompts versioned in Helicone need recreating in Opik's Prompt Library. Scores you computed outside Helicone and posted to it can become built-in or custom metrics that Opik runs itself.
A workable order is to move tracing first while Helicone keeps routing traffic, then set up Opik's online rules and annotation queues on the new traces, then replace the gateway. Teams that self-host Helicone in the meantime should keep up with its patches: September 2026 fixed a platform-admin takeover and a cross-tenant read in its query language, and October brought further hardening for self-hosted deployments.
Helicone's pricing page still lists, as of 20 October 2026, a free Hobby plan with 10K requests, 1 GB of storage, 1 seat and 7-day retention, Pro at $79 a month and Team at $799 a month. With sign-ups disabled, new customers cannot start on any of them. Opik's Free Cloud covers 25k spans a month for up to 10 team members with 60-day retention, and Pro Cloud is $19 a month for 100k spans, then $5 per 100k. Both can be self-hosted for free. See Helicone's pricing and Opik's pricing.
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 a failing score through runtime policies, approval holds and a kill switch for a run, an agent or a team. Helicone's gateway blocks prompt injection for OpenAI models only, and Opik's guardrails depend on calls in your code; neither has approval holds or tool-call policies. Prefactor has no prompt playground or prompt registry, no gateway, and its platform is hosted, so it does not replace Helicone's routing. 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.