Best LLM Observability Tools

As of July 28, 2026, Langfuse is our pick for teams that need LLM traces, datasets, evaluations, and an open-source self-hosting route in one product; LangSmith wins when LangChain or LangGraph already anchors the application. Phoenix fits private OpenTelemetry deployments, Braintrust fits evaluation-led development, and Helicone fits gateway-first request monitoring. We reached that verdict by reviewing current first-party documentation, not by running the tools. This article does not rank trace latency, ingestion throughput, evaluator accuracy, or support quality.

This is the observability and evaluation layer of the AI App Stack. There are no partner or sponsored links in this roundup. Inclusion and order come from the documented product fit as it stood on the review date.

Start with the failure you need to catch

A trace answers what happened. An evaluation answers whether the result was acceptable. A dataset makes that judgment repeatable, and an experiment tells you whether a prompt, model, retrieval step, or tool change improved the same cases. Buying a trace viewer without planning that loop leaves you with a precise record of bad answers.

We checked five parts of each product's current first-party documentation:

Table 1 Download CSV

Decision	What we looked for	Why it changes the pick
Trace model	Nested model, retrieval, tool, and custom spans	A flat request log cannot explain an agent path
Evaluation loop	Offline datasets, production scores, human review, and experiment comparison	Visibility alone does not catch a quality regression
Instrumentation	SDKs, framework hooks, proxy paths, and OpenTelemetry support	Portable instrumentation lowers switching cost
Data boundary	Hosted, open-source, hybrid, or self-hosted deployment	Prompts and tool results can contain sensitive data
Cost shape	Seats, traces, observations, requests, data, scores, and infrastructure	The cheap starting plan may not match production volume

The sources establish available features and published limits. They do not establish which interface is easier under incident pressure, which collector drops fewer spans at load, or which evaluator agrees most often with a domain expert. We would need the same instrumented application and labeled cases inside every product to make those claims.

The map is a selection aid from published product boundaries, not a benchmark result. Start with the branch that can disqualify a tool, such as full open-source control or an existing LangGraph stack, before comparing dashboard details.

The documentation comparison

Table 2 Download CSV

Tool	Best for	Ingestion path	Evaluation loop	Deployment boundary	Cost shape	Free starting route	Where it loses
Langfuse	Integrated tracing, prompts, datasets, and evals with cloud-to-OSS portability	OpenTelemetry-based Python and JS/TS SDKs; OTLP	Offline experiments plus code, model, API, and human scores	Cloud or MIT-licensed self-hosted core	Traces + observations + scores as billable units in Cloud	Hobby Cloud plan or self-hosted OSS	Version compatibility and self-hosted operations need active ownership
LangSmith	LangChain and LangGraph applications	Framework integrations, manual tracing, SDKs, and OpenTelemetry	Versioned datasets, offline experiments, and online evaluators	SaaS; enterprise self-hosted and hybrid options	Seat plus tracing and platform usage	Single-seat Developer plan	Less attractive when framework neutrality or open-source control is the priority
Phoenix	OpenTelemetry-first teams that want the full data path in their infrastructure	OTLP with OpenInference instrumentation	Code, model, and human evals over traces, datasets, and experiments	Open-source Docker, Kubernetes, or cloud deployment	Your infrastructure for OSS; separate hosted Arize plans	Self-hosted open source	One Phoenix instance is one tenant; production operations remain yours
Braintrust	A production-to-evaluation feedback loop	Braintrust SDKs and an OTLP endpoint	Logs and experiments share a structure; offline and online scoring	Hosted or customer-run data plane with a managed control plane	Processed data, scores, and plan features	Starter plan	The self-hosted option does not make the whole control plane open source or customer-run
Helicone	Request, session, cost, user, and gateway monitoring	AI gateway, proxy, SDK, or logging integrations	Datasets and externally produced scores	Hosted plans; enterprise on-premises option	Plan plus requests and storage	Hobby plan	Helicone documents that it stores scores but does not run evaluations for you

We kept exact dollar figures and quotas out of the table because they change faster than the product boundary. Follow the linked Langfuse, LangSmith, Arize, Braintrust, and Helicone pricing pages for the current numbers.

Langfuse owns the integrated open path

Langfuse puts tracing, prompt management, datasets, experiments, and several evaluation methods in one product. Its evaluation concepts connect offline experiments to online scoring: a production failure can become a dataset item, and a later experiment can run the revised application against that case. Code evaluators, LLM-as-a-judge, API scores, UI scores, and annotation queues all write into the score model.

The deployment boundary is the reason it leads this list. Langfuse publishes its core product under the MIT license and documents self-hosted pricing separately from Cloud. Cloud billing counts traces, observations, and scores as billable units, which is clearer than pretending one request always equals one stored object.

Where it loses: open source removes a license bill, not an operations bill. A production deployment still needs databases, storage, upgrades, backups, and someone who reads compatibility notes. That last part matters in July 2026. The current version compatibility page marks Cloud v4 as generally available while saying the self-hosted v4 schedule will follow. A team adopting current SDK majors should check that matrix against its server before an upgrade.

Langfuse is our default shortlist when the team wants one evaluation loop and a credible route between managed Cloud and self-hosting. If the application is already deep in LangChain or the team will not operate observability infrastructure, the next two choices can be cleaner.

LangSmith fits a LangChain-native team

LangSmith has the shortest documented route from LangChain or LangGraph code to a useful trace. Its observability model groups runs into traces and supports framework integrations, manual instrumentation, tags, and metadata. The same platform adds versioned datasets and both offline and online evaluation.

That integration changes the amount of glue code a team owns. A failing production run can move into a dataset, receive code or model-based evaluators, and appear beside later experiments. OpenTelemetry ingestion is also available, so LangSmith is not limited to LangChain spans, but the tightest workflow still appears when the application already uses the LangChain product family.

Where it loses: the hosted starting path combines plan, seat, and usage decisions, while self-hosted LangSmith is an enterprise deployment with PostgreSQL, Redis, ClickHouse, and supporting services. It is not the same proposition as installing an MIT-licensed tool and owning every layer. For a small framework-neutral service, that can be more platform than the job needs.

Choose LangSmith when existing LangChain or LangGraph integration removes work you would otherwise build. Do not choose it merely because an agent tutorial already has the tracing environment variable in its final step.

Phoenix keeps the stack in your infrastructure

Phoenix is built on OpenTelemetry and OpenInference, and its current product overview covers traces, evaluations, prompt work, datasets, and experiments. It accepts model, retrieval, tool, and custom spans, then lets code, model-based, or human evaluators attach scores to traces and experiment results.

Its self-hosted shape is unusually concrete. The deployment architecture describes a web application, OTLP collector, and SQL backend. SQLite fits a local single-user start; PostgreSQL is the production option. Docker and Kubernetes deployment guides are public, and the privacy documentation says trace and evaluation data stays inside the operator's infrastructure when Phoenix is self-hosted.

Where it loses: a Phoenix instance represents one tenant. The current architecture recommends separate instances or database schemas for stronger team and environment isolation, so a large organization must design that boundary. Continuous monitoring and alerting also extend into the broader Arize AX product line rather than every feature living in Phoenix OSS.

Phoenix is our pick when keeping trace data in a private OpenTelemetry stack is the first requirement and the team is prepared to run it. It is not a free managed service disguised as a container.

Braintrust starts from the evaluation loop

Braintrust treats production observation and pre-release evaluation as one data problem. Its observability documentation says logs use the same structure as experiments, so teams can filter production traces, apply feedback, and turn selected failures into targeted datasets. The evaluation workflow covers offline experiments, online scoring, playground work, and regression checks.

This is the strongest fit when the organization's main question is not "where did the tokens go?" but "did this change improve the task?" Braintrust supports SDK instrumentation and OpenTelemetry export, which makes it possible to keep a broader trace graph while using the same spans in evaluation work.

Where it loses: its self-hosted deployment is a customer-run data plane, not a fully customer-run product. Sensitive traces, datasets, prompts, and experiment data stay in that plane, while Braintrust continues to provide the UI, authentication, metadata, and platform updates through its control plane. That split can satisfy a data-residency requirement, but it does not satisfy every full-isolation or open-source requirement.

Braintrust belongs on the shortlist when experiments and production feedback already drive releases. If the team mainly needs a request gateway and cost report, it starts too far down the evaluation stack.

Helicone starts from the request path

Helicone's center of gravity is the request. Its gateway and logging paths expose models, tokens, costs, errors, users, custom properties, and sessions. The sessions documentation groups LLM calls, vector queries, tools, and other logged requests into a trace-like hierarchy, which is useful for multi-step agents that already pass through its gateway.

It also supports datasets and score storage, but the boundary is explicit. The current score documentation says Helicone does not run evaluations; teams submit results from an evaluation framework or their own workers. Its older prompt Experiments feature was marked for removal in 2025. A team choosing Helicone for an evaluation lab should budget another system for the actual judges, test execution, and experiment comparison.

Where it wins is operational simplicity for gateway-first monitoring. A team that wants per-user cost, request filtering, rate limits, caching, provider fallbacks, and session inspection can get those concerns close to the API path. Where it loses is the extra seam between production traces and repeatable evals.

Use Helicone when the gateway is the product boundary you want to observe. Do not infer an evaluator from a score column.

Pick by scenario, then run a bake-off

Table 3 Download CSV

Scenario	First tool to trial	The disqualifying test
One integrated cloud and open-source path	Langfuse	Confirm the server and SDK versions match, then replay one dataset experiment
LangChain or LangGraph already owns the agent loop	LangSmith	Trace non-framework code and verify the dataset handoff is still clean
Trace data must stay inside your infrastructure	Phoenix	Run the intended PostgreSQL and authentication setup, not the local SQLite demo
Releases are governed by datasets and experiments	Braintrust	Turn a real production failure into a scored regression case
The gateway must own cost, routing, and request analytics	Helicone	Export one failure into the external evaluation system you will actually use
Existing OpenTelemetry APM already answers the incident question	Keep it, then add one specialist only if needed	Prove that the new product closes an evaluation or LLM-semantics gap

We would compare two finalists with the same small workload rather than five products with five demo applications. Instrument one representative RAG or agent path with model, retrieval, tool, and application spans. Include a successful case, a provider error, a bad retrieval, a malformed tool call, a slow response, and a prompt that contains sensitive text that must be redacted.

Then run this sequence:

1. Send the same trace shape to both products and check whether parent-child relationships survive.
2. Find the bad retrieval without knowing its trace ID.
3. Convert that failure into a labeled dataset case.
4. Run one code-based evaluator and one human review pass.
5. Compare a changed prompt against the original cases.
6. Export the data, estimate the production billing unit, and rehearse deletion.

Record setup time, missing spans, time to isolate the failure, evaluator reproducibility, redaction behavior, data export, and monthly cost at the expected volume. Do not use the vendor's sample chatbot as the test workload. It has already been arranged to make the vendor's happy path happy.

This article has not run that bake-off. Until it does, the verdict is about documented workflow and deployment fit, not runtime quality.

Where observability sits in the stack

Observability begins before deployment. The RAG model guide helps choose the generation layer, but a production trace must also show retrieved document identifiers, ranking or filtering steps, tool calls, and the final answer. Otherwise a groundedness failure looks like a model failure even when retrieval supplied the wrong evidence.

The deployment guide covers runtime boundaries such as secrets, timeouts, streaming, and background work. Add the trace exporter to that design before traffic arrives, and make sure it flushes under the actual serverless or worker lifecycle. A local span that never leaves the process is comforting but not observable.

Model tokens remain a separate cost layer. Use the LLM pricing comparison for provider rates, then add observability ingestion, evaluator model calls, retained data, and the people reviewing failures. The cheapest trace store can become the expensive choice if nobody can turn a production miss into a regression test.

The trace store is the commodity here; the loop from production failure to repeatable test is the product, and it belongs to whichever tool your team will actually close it in.