Both are open-source tools you can self-host for LLM logging, cost tracking and prompt versioning. Helicone has been in maintenance mode since March 2026 and is not taking new cloud sign-ups; Langfuse is still shipping under ClickHouse and adds offline and online evaluation. For most readers this page is about moving from one to the other.[3][6]
Helicone and Langfuse are both open source and self-hostable, but Helicone is a request logger and AI gateway in maintenance mode since March 2026, with new cloud sign-ups disabled, while Langfuse is an active tracing, evaluation and prompt versioning platform. ClickHouse acquired Langfuse in January 2026 and Langfuse says its licence is unchanged. On 19 October 2026 Helicone said it is no longer part of Mintlify, with no new owner named. For a new deployment Langfuse is usually the practical choice.
| Decision factor | Helicone | Langfuse |
|---|---|---|
| Ownership | Acquired by Mintlify in March 2026; on 19 October 2026 said it is no longer part of Mintlify, no new owner named | Acquired by ClickHouse in January 2026; licence and self-hosting unchanged, per Langfuse |
| Status | Maintenance mode; cloud sign-ups disabled since March 2026 | Active; release v4.64.0 on 19 October 2026 |
| Licence | Apache-2.0; separate Rust gateway GPL-3.0, last pushed November 2025 | MIT, except add-ons in the ee folders |
| Evaluation | Not an evaluation framework, by its own docs | Experiments, evaluator rules on live traffic, LLM and code evaluators, annotation queues |
| Agent tracing | Header-based session grouping | Nested observations, agent graph view, sessions |
| Gateway | AI Gateway for 100+ providers with failover, caching and rate limits | None; ingests traces from LiteLLM Proxy and Vercel AI Gateway |
| Free tier | Hobby: 10K requests, 7-day retention, existing accounts only | Hobby: 50k units a month, 30 days of data |
| First paid tier | Pro: $79 a month, existing accounts only | Core: $29 a month, 100k units |
| Capability | Helicone | Langfuse | 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 | — | ✓ |
| Gateway or proxy in the request path | ✓ | — | — |
| Licence and deployment | |||
| Open source | ✓ | ✓ | Partial |
| Self-hosting | ✓ | ✓ | Partial |
Mintlify announced its acquisition of Helicone on 3 March 2026 and put the product into maintenance: patches, bug fixes and new model support, with help for customers moving elsewhere. Cloud sign-ups were disabled that day and the changelog shows no feature entries after November 2025. On 19 October 2026 the Helicone site began saying it is no longer part of Mintlify, with details to come.
ClickHouse acquired Langfuse in January 2026. Langfuse said its roadmap would not change, its licence would not change and self-hosting would remain a first-class path, and it has kept shipping: code evaluators in May, a 1.0 CLI in August and evaluator backfills in September. One date for existing Langfuse Cloud users to note: Cloud becomes v4-only on 16 November 2026.
Helicone also had a four-day outage of its dashboard and log processing in May 2026 when AWS locked its account, while the proxy kept running.
Helicone's main repository is Apache-2.0 and self-hosts with Docker, Docker Compose or Helm. Its Docker image no longer proxies LLM traffic, so the AI Gateway has to be deployed alongside it, and that Rust gateway is a separate GPL-3.0 repository last pushed in November 2025. Helicone logs requests, groups them into sessions by header, tracks cost and versions prompts that deploy through the gateway. Its docs say it does not run evaluations, and its Python evaluators were turned off in October 2026.
Langfuse is MIT-licensed apart from add-ons in its ee folders, and self-hosts on Docker Compose for testing or Kubernetes for production, with Postgres, ClickHouse, Redis and object storage. It traces nested agent steps, runs experiments on datasets or production traces, scores live traffic with evaluator rules, and queues traces for human review. It is not a proxy: spans are exported out of the request path.
Cost tracking is close. Helicone computes exact per-request cost through its gateway's model registry and alerts on graduated thresholds. Langfuse records token usage and cost per generation, ingested or inferred from model definitions, with custom model prices and per-user cost. Langfuse Cloud runs in EU, US, Japan and HIPAA regions; Helicone's cloud lists US and EU regions.
Start with instrumentation. Helicone's async SDKs are built on OpenLLMetry, and Langfuse names OpenLLMetry as compatible with its OTLP endpoint, which accepts HTTP but not gRPC. Teams that log through Helicone's proxy will need Langfuse's Python or TypeScript SDK, a framework integration, or OpenTelemetry instead.
Then replace the gateway, if you use it. Langfuse has none, but it ingests traces from third-party gateways such as LiteLLM Proxy and Vercel AI Gateway, so one of those can take over routing while Langfuse records the calls. Prompts move to Langfuse's prompt versioning, which deploys versions by label without code changes. Self-hosted Helicone keeps running under Apache-2.0 while you switch, with maintenance-only updates.
Feedback moves too. Helicone attaches positive or negative feedback to requests and sessions through its API; Langfuse collects user feedback through its browser SDK, server SDKs or API. Once traces arrive, Langfuse experiments can run against production traces without a dataset.
As of 20 October 2026, Langfuse Hobby is free for 50k units a month with 30 days of data, and Core is $29 a month for 100k units, then $8 per 100k; self-hosting is free. Helicone lists Hobby free with 10K requests, Pro at $79 a month and Team at $799 a month, available only to existing accounts while sign-ups are disabled. Self-hosting either removes the subscription and adds the infrastructure you run. See Helicone's pricing and Langfuse's pricing.
Langfuse Pro is $199 a month with three years of data access, and Enterprise is $2,499 a month or custom. Above 50M units the Core overage rate falls to $6 per 100k.
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 acts on what it finds through runtime policies on tool calls and data access, approval holds for risky actions, and a kill switch for a run, an agent or a team. Langfuse evaluates after the fact and does not block, and Helicone's gateway can block prompt injection for OpenAI models only. Prefactor has no prompt playground or prompt registry, no gateway, and its platform is hosted, so if free self-hosting is the requirement, Langfuse is the better pick. 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.