PrefactorvsLangfusevsLangSmith

Langfuse vs LangSmith: how to choose

Feature lists now overlap almost completely. Langfuse, now owned by ClickHouse, is the open-source option you can self-host for free on any framework; LangSmith is the first-party choice for teams on LangChain or LangGraph, with hosted extras such as Engine, agent deployment and a model-call gateway.[3][4]

Book a demo →View docs

support-agent v4 · one run, two layersexample

Illustrative run, showing what each layer tells you

Langfuse / LangSmithsee

fetch_customer212ms · 1.2k tok

apply_refund1.4s · 3.1k tok

send_reply340ms · 0.8k tok

trace recorded, no verdict

Prefactoradds

Did its job✓ yes

Quality84 / 100

Cost$0.42 · in budget

Drift vs baselinenone

✓ a run that breaches its schema is held for review

§01 / THE SHORT ANSWERtl;dr: which, and when

TL;DR

What is the difference between Langfuse and LangSmith?

Langfuse is open source (MIT, apart from licensed add-ons) and free to self-host on any framework, while LangSmith is LangChain's proprietary platform with the deepest LangChain and LangGraph integration and self-hosting only on Enterprise. ClickHouse acquired Langfuse in January 2026, and Langfuse says its roadmap, licence and self-hosting are unchanged. Both trace agents, run offline and online evals, queue runs for review and version prompts.

The short answer

At a glance (checked 20 October 2026)

Decision factor	Langfuse	LangSmith
Ownership	Acquired by ClickHouse, January 2026; roadmap and licence unchanged, per Langfuse	LangChain, independent; $125M Series B in October 2025
Licence	MIT, except add-on features in the `ee` folders	Proprietary platform; client SDKs are MIT
Self-hosting	Free, via Docker or Kubernetes	Enterprise add-on on Kubernetes; BYOC also Enterprise
Framework coupling	Framework-neutral SDKs and OpenTelemetry	Works standalone; deepest with LangChain, LangGraph and Deep Agents
OpenTelemetry	OTLP over HTTP; gRPC not supported	OTLP ingest; maps OpenInference and OpenLLMetry; SDK can export to other backends
Run-time controls	None; evaluation runs after the fact	Partial: LLM Gateway (beta) caps spend, restricts models and redacts model calls
Free tier	Hobby: 50k units a month, 30 days of data; self-hosting free	Developer: 1 seat, 5k base traces a month, 14-day retention
First paid tier	Core: $29 a month, 100k units, then $8 per 100k	Plus: $39 per seat a month, 10k base traces; overage rate not published

§02 / CAPABILITY MATRIXside by side: what each covers

Side by side

Langfuse vs LangSmith: capabilities side by side

Capability	Langfuse	LangSmith	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	—	Partial	✓
Gateway or proxy in the request path	—	—	—
Licence and deployment
Open source	✓	Partial	Partial
Self-hosting	✓	✓	Partial

§03 / IN DEPTHdetail: the reasoning

In depth

Who runs it, and who owns the code

Langfuse is MIT-licensed with a free self-hosted path, while LangSmith is proprietary and self-hosts only as an Enterprise add-on. Langfuse runs on Docker Compose or Kubernetes with Postgres, ClickHouse, Redis and object storage, apart from a few licence-keyed ee add-ons, so you pay in infrastructure and engineering time instead of a contract.

LangSmith self-hosts with Helm or Terraform, and BYOC is Enterprise only too. LangSmith's hosted service offers US, EU and APAC data residency; Langfuse Cloud offers EU, US, Japan and HIPAA regions.

What the ClickHouse acquisition changed

Little so far: ClickHouse acquired Langfuse on 16 January 2026, and Langfuse said its roadmap, licensing and self-hosting would not change. It has kept shipping, and the latest release, v4.64.0, came out on 19 October 2026.

This year's releases include experiments as a top-level feature, code evaluators in Python and TypeScript, a GitHub Action that fails CI below a score threshold and evaluator backfills over up to six months. Plan for one change: Langfuse Cloud becomes v4-only on 16 November 2026, which may mean migration work for existing cloud users.

If you build on LangChain or LangGraph

LangSmith's advantage is first-party depth on LangChain and LangGraph; off that stack it narrows to roughly parity on tracing and evals. Its Trajectory view, open to all users since September 2026, combines messages across a main agent and its subagents.

LangSmith also offers Engine, in public beta, which drafts fixes from production traces, and LangSmith Deployment for LangGraph agents. Langfuse traces LangChain and LangGraph too, among many frameworks, with an agent graph view since July 2026. Neither blocks a bad tool call: Langfuse leaves run-time protection to separate libraries, and LangSmith's LLM Gateway, in public beta since 30 July 2026, can cap spend, restrict models and redact personal data in model calls, with tool and MCP gateways planned.

Pricing shape

Langfuse charges per unit ingested and LangSmith per seat plus traces, and LangSmith's per-trace overage rate is no longer public, so high-volume costs cannot be compared from published figures. As of 20 October 2026, Langfuse Hobby is free for 50k units a month, Core is $29 a month for 100k units, then $8 per 100k, falling to $6 above 50M, Pro is $199 a month and Enterprise starts at $2,499 a month.

Traces, observations and scores each count as units, so an agent run usually emits several. LangSmith Developer is free for one seat and 5k base traces, and Plus is $39 per seat a month with 10k base traces. Model your own volumes against Langfuse's pricing and LangSmith's pricing.

§04 / THE QUALITY GAPour take: where it stops

Our take

Where Prefactor fits

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, not only the final answer, 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 has no run-time controls, and LangSmith's gateway covers model calls only. Prefactor has no prompt playground or prompt registry and its platform is hosted, so if free self-hosting or prompt iteration is what you need, Langfuse or LangSmith is the better pick. See the full comparison of 15 platforms.

See it on your own agents

A working session on a fleet like yours: watch a run evaluated, catch a drift, walk the record.

Book a demo →View docs

§05 / WHICH TO PICKdecide: by your stack

Which to pick

Which one fits

Pick Langfuse if

You need to self-host without an Enterprise contract.
You want an MIT-licensed codebase you can inspect and run.
You use several frameworks, or none, and want neutral instrumentation.
A low first paid tier, or no subscription at all, matters most.

Pick LangSmith if

You build on LangChain, LangGraph or Deep Agents.
You want hosted extras: Engine, agent deployment and a model-call gateway.
Hosted-only is acceptable, or an Enterprise budget exists for self-hosting.
Per-seat pricing fits your team size.

§06 / HOW WE REVIEWEDsources: checked October 20, 2026

Methodology

How we reviewed this comparison

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.

Sources reviewed

Prefactor context

Methodology

Reviewed public product, documentation, and launch material visible at the time of writing.
Mapped each page to the primary buyer, control layer, and runtime capabilities each vendor describes publicly.
Prefer direct product and documentation pages over analyst summaries or reseller material.

§07 / QUESTIONSfaq: the common ones

Questions

What is the difference between Langfuse and LangSmith?

Both trace LLM and agent applications, run offline and online evals, queue runs for review and version prompts. Langfuse is open source and free to self-host on any framework. LangSmith is LangChain's proprietary platform, with the deepest LangChain and LangGraph integration and self-hosting on Enterprise only.

How does Prefactor compare with Langfuse and LangSmith?

Prefactor is an agent evaluation platform for teams building AI agents: it observes every agent run from development to production, evaluates whether the agent did its job, and shows how to improve it. As of 20 October 2026, Prefactor can block or hold a live agent action; LangSmith does so in part and Langfuse does not. Langfuse and LangSmith both include prompt versioning and a playground. Prefactor evaluates how prompts perform in live runs rather than providing a prompt playground or registry.

Is LangSmith only for LangChain?

No. LangSmith traces other applications through its SDK, wrappers for OpenAI and Anthropic, integrations with frameworks such as CrewAI, Pydantic AI and the Vercel AI SDK, and an OpenTelemetry endpoint. Its advantage is largest for teams already on LangChain or LangGraph.

Which is better for agents?

Both trace multi-step agents. LangSmith has the deepest view of LangGraph agents and a Trajectory view across subagents; Langfuse traces agents from any framework, has an agent graph view and lets you self-host. The deciding factor is usually your framework and whether you need to self-host.

Can I move from LangSmith to Langfuse, or the other way?

Tracing moves fairly easily, since both use OpenTelemetry: LangSmith's SDK can export to other OTel backends, and Langfuse accepts OTLP over HTTP. Datasets, evaluators and prompts do not move automatically and need to be exported and rebuilt.

Did anything change for Langfuse after the ClickHouse acquisition?

Not in licence or self-hosting, as of 20 October 2026. Langfuse said its roadmap and licence would not change and has kept shipping. Separately, Langfuse Cloud becomes v4-only on 16 November 2026, which may need migration work for existing cloud users.

Who owns Langfuse?

As of 20 October 2026: ClickHouse, Inc., which acquired Langfuse in January 2026. Langfuse said its roadmap and licence would not change and that self-hosting stays a first-class option.

Is Langfuse open source?

Yes. The repository is MIT-licensed, apart from add-on features in its ee folders that need a licence key. Self-hosting on Docker or Kubernetes is free.

How much does Langfuse cost?

As of 20 October 2026: Hobby is free with 50k units a month and 30 days of data, and self-hosting is free. Core is $29 a month for 100k units, then $8 per 100k. Pro is $199 a month, and Enterprise is $2,499 a month or custom.

Can Langfuse block or hold a live agent action?

No. Langfuse evaluates runs after they happen. Its documentation says run-time protection comes from separate libraries, and Langfuse scores those measures afterwards.

Who owns LangSmith?

As of 20 October 2026: LangChain, Inc., a privately held company that raised a $125M Series B at a $1.25B valuation in October 2025. No acquisition has been announced.

Is LangSmith open source?

Partly. The client SDKs are MIT-licensed and the LangChain frameworks are open source, but the LangSmith platform is proprietary. Self-hosting requires an Enterprise contract.

How much does LangSmith cost?

As of 20 October 2026: Developer is free for one seat and 5k base traces a month. Plus is $39 per seat a month with 10k base traces; the overage rate is not published. Enterprise is a custom annual contract.

Can LangSmith block or hold a live agent action?

Partly. The LLM Gateway, in public beta since July 2026, can cap spend, restrict models and redact personal data in model calls. It does not cover tool calls; LangChain describes tool and MCP gateways as planned.

Is Prefactor open source?

Partly. Its deterministic evals package is Apache-2.0; the platform is hosted.

How much does Prefactor cost?

As of 20 October 2026: Dev is free with 5,000 spans a month. Startup is $49 a month and Scaleup $199 a month. Enterprise is custom above 4 million spans a month.

Can Prefactor block or hold a live agent action?

Yes. Runtime policies can block, throttle, sandbox or escalate an agent's tool calls and data access before they run, approval routing holds risky actions for a person, and a kill switch stops a run, an agent or a team.

Reviewed against public sources on October 20, 2026Suggest a correction

Score the run, then act on it

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.

Agent Performance Platform
Unified performance platform for agents, authentication, and risk management

All Systems Operational

3Global Agents

7Instances

5Services

12%Human Intervene

4High Risk

$2,360Monthly Spend

Mission ControlLive agent health with 7-day activity heartbeat

Claims Proc...68

$330/moRed

Claims Proc...65

$160/moRed

Claims Proc...82

$170/moAmber

ChatGPT74

$150/moAmber

Critical Alerts

Pending Review

Resolved Today

Total Actions

Operational Actions3

Access changes, policy reviews, workflow drift

Risk Actions5

Sensitive data exposure, unsafe actions, access violations

Unauthorized access to financial database

Riskcriticalresolved

Agent attempted to access Finance-MCP server without proper authorization.

Claims Processor v1.0Finance-MCP03/02/2025, 14:32:00

Action taken by Prefactor:

Prefactor revoked the agent’s active session and blocked further access attempts.

✅ Resolved by Security Team at 03/02/2025, 14:45:00

Event LogSecurityPermissions

Mcp CallEVT-001 • 02/10/2025, 17:30:22Message sent to Slack channel200ms · $0.0500

PromptEVT-002 • 02/10/2025, 17:25:15User prompt received337ms · $0.1200

Mcp CallEVT-003 • 02/10/2025, 17:26:42Retrieved repository information474ms · $0.0800

Tool CallEVT-004 • 02/10/2025, 17:28:10Analyzed code for security issues611ms · $0.1500

OutcomeEVT-005 • 02/10/2025, 17:29:55Analysis complete748ms · $0.0300

Langfuse vs LangSmith: how to choose

What is the difference between Langfuse and LangSmith?

At a glance (checked 20 October 2026)

Langfuse vs LangSmith: capabilities side by side

Who runs it, and who owns the code

What the ClickHouse acquisition changed

If you build on LangChain or LangGraph

Pricing shape

Where Prefactor fits

See it on your own agents

Which one fits

Pick Langfuse if

Pick LangSmith if

How we reviewed this comparison

Sources reviewed

Methodology

Related

Score the run, then act on it

See how every agent performs, and make it better