Both trace agents, run experiments and online evals, queue traces for human review and version prompts. Langfuse is the open-source option you can run yourself for free, with online evals included; Maxim AI is hosted and proprietary, built around simulating text and voice agents, with its own gateway. Whether you must self-host decides most cases before features do.[3][4]
Langfuse is open source (MIT, apart from licensed add-ons) and free to self-host, including online evals; Maxim AI is a proprietary, hosted platform for simulating, evaluating and tracing agents, with an Apache-2.0 gateway, Bifrost. ClickHouse acquired Langfuse in January 2026 and Langfuse says its licence and self-hosting are unchanged. Maxim AI adds voice simulation and a gateway; Langfuse adds free self-hosting and longer free retention.
| Decision factor | Langfuse | Maxim AI |
|---|---|---|
| Ownership | Acquired by ClickHouse in January 2026; roadmap and licence unchanged, per Langfuse | Independent; $3M seed led by Elevation Capital in 2024 |
| Licence | MIT, except add-on features in the ee folders | Platform proprietary; Bifrost gateway and SDKs Apache-2.0 |
| Self-hosting | Free, via Docker or Kubernetes | Enterprise in-VPC only; Bifrost self-hosts free |
| Pre-release testing | Experiments on datasets, production traces or local data | Multi-turn text and voice simulation, plus test runs |
| Gateway | None; traces are exported out of the request path | Bifrost: failover, caching, virtual keys, MCP gateway |
| Runtime blocking | None; evaluation happens after the fact | Through Bifrost Enterprise guardrails |
| Free hosted tier | Hobby: 50k units a month, 30 days of data | Developer: up to 3 seats, 10k logs a month, 3-day retention |
| First paid tier | Core: $29 a month, 100k units | Professional: $29 per seat a month, 100k logs |
| Capability | Langfuse | Maxim AI | Prefactor |
|---|---|---|---|
| Observe | |||
| LLM call tracing | ✓ | ✓ | ✓ |
| Multi-step agent tracing (tool calls, nested spans) | ✓ | ✓ | ✓ |
| Cost tracking | ✓ | ✓ | ✓ |
| OpenTelemetry support | ✓ | ✓ | ✓ |
| Evaluate | |||
| Offline evals (datasets, experiments) | ✓ | ✓ | ✓ |
| Online evals on production traffic | ✓ | ✓ | ✓ |
| LLM-as-a-judge scoring | ✓ | ✓ | ✓ |
| Code-based evaluators | ✓ | ✓ | ✓ |
| Human review and annotation | ✓ | ✓ | ✓ |
| Prompt versioning and playground | ✓ | ✓ | — |
| Act on findings | |||
| Runtime controls on live agent actions | — | ✓ | ✓ |
| Gateway or proxy in the request path | — | ✓ | — |
| Licence and deployment | |||
| Open source | ✓ | Partial | Partial |
| Self-hosting | ✓ | ✓ | Partial |
Langfuse publishes its platform under the MIT licence, apart from a set of add-on features in licence-keyed ee folders, and documents free self-hosting on Docker Compose or Kubernetes, backed by Postgres, ClickHouse, Redis and object storage. The free self-hosted version includes tracing, experiments, prompt versioning and online evaluator rules on production traffic. ClickHouse acquired Langfuse in January 2026; Langfuse said its roadmap and licence would not change and self-hosting would stay first-class, and it has shipped steadily since, including code evaluators in May and evaluator backfills in September.
Maxim AI's evaluation platform is proprietary. Self-hosting it means an Enterprise in-VPC deployment on AWS, GCP or Azure, either fully isolated or as a data plane peered to Maxim's control plane. Its open-source parts are the SDKs and Bifrost, the gateway. If your traces cannot leave your network and you have no budget for an Enterprise contract, Langfuse is the option that works.
One change to plan for: Langfuse Cloud becomes v4-only on 16 November 2026, which may need migration work for existing hosted users.
Langfuse starts from the trace. You instrument, watch production, then build datasets and experiments from what you see; since April 2026 experiments can run against dataset items, production traces or local data, and a GitHub Action fails CI below a score threshold. Its agent graph view arrived in July 2026.
Maxim AI starts before production. It simulates multi-turn conversations with text and voice agents, runs test runs for prompts, no-code agents and agents behind an HTTP endpoint, and lists simulation for AWS Bedrock Agents and Glean agents. Human evaluators get rating guidelines, and external raters can be added by email. If your agent speaks to customers by voice, this is the larger difference between the two.
Evaluator options are close. Langfuse runs LLM-as-a-judge evaluators, Python or TypeScript code evaluators written in the UI since May 2026, a low-cost Jev-as-a-judge classifier since September, and annotation queues. Maxim AI offers pre-built evaluators, custom LLM judges with a chosen model and scale, JavaScript or Python code evaluators, and API-based evaluators that call your own endpoint.
Langfuse is not a proxy and is never required in the request path; spans are exported asynchronously from SDKs or OpenTelemetry. It does not block anything at run time, and its documentation says protection comes from separate libraries, with Langfuse evaluating those measures afterwards. Its threshold alerts notify rather than block.
Maxim AI's Bifrost sits in the request path by design: an OpenAI-compatible gateway with failover, load balancing, semantic caching, virtual keys with budgets and an MCP gateway. Its guardrails check inputs, outputs and MCP tool-call arguments and results and can block, redact or transform content, as a Bifrost Enterprise feature. Langfuse can ingest traces from third-party gateways such as LiteLLM Proxy, so a team that wants Langfuse plus a gateway can still pair them.
As of 20 October 2026, Langfuse charges per unit ingested (traces, observations and scores). Hobby is free for 50k units a month with 30 days of data, Core is $29 a month for 100k units and then $8 per 100k, Pro is $199 a month and Enterprise $2,499 a month or custom. Self-hosting is free. Maxim AI's Developer tier is free for up to 3 seats and 10k logs a month with 3-day retention; Professional is $29 per seat a month with 100k logs and Business $49 per seat a month with 500k logs, with overage at $1 per 10k logs. Maxim AI lists online evals from Professional upwards.
At one seat the entry prices match; from two seats up, Langfuse Core costs less before usage. Retention also differs: Maxim AI keeps free-tier logs for 3 days and Professional logs for 7, against 30 and 90 days on Langfuse Hobby and Core. Check Langfuse's pricing and Maxim AI's pricing against your own volumes.
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, approval holds and a kill switch scoped to a run, an agent or a team. Langfuse evaluates after the fact; Maxim AI can block content through Bifrost Enterprise guardrails but documents no approval holds. Prefactor has no prompt playground or prompt registry and its platform is hosted, so if free self-hosting is the requirement Langfuse is the better pick, and for voice simulation Maxim AI is. 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.