The Best Self-Hosted LLM Observability Tools in 2026, Ranked by License and Ops Reality
Four platforms you can run on your own infrastructure - judged on license honesty, how complete the self-host actually is, the ops burden, and maturity. Where "self-host" means the real product, and where the license has an asterisk.
Published:
The reason to self-host observability is simple - your prompts, outputs and traces never leave your infrastructure. But “self-hostable” hides a lot. Some tools give you the real product under a permissive license. Some give you a hobbled core and gate the good parts behind a cloud plan. Some ship a license that reads open but is not. I ranked these four on the two things that actually decide a self-host - what license you get, and how much ops the self-host really costs.
The axes:
- License honesty. MIT and Apache-2.0 are OSI open source. Elastic License 2.0 is source-available - narrower freedom.
- Self-host completeness. Do you keep the real features, or a stripped core?
- Ops burden. How many services you stand up and keep running.
- Maturity. Because a self-hosted stack is a multi-year bet.
Here is the ranked list.
The short version
| Tool | License | Self-host completeness | Ops burden | Maturity |
|---|---|---|---|---|
| Langfuse | MIT | Near-complete, 3 gated | Four services | High |
| Opik | Apache-2.0 | Full, zero gates | Kubernetes-scale | High (Comet) |
| Laminar | Open-source | Full stack, agent-focused | Whole stack, Rust | Youngest |
| Arize Phoenix | ELv2 server | No feature gates | Sub-minute local start | Well-funded |
1. Langfuse - the battle-tested self-host default
Langfuse is the self-host default because it earns it. It is MIT-licensed, and only three features are enterprise-gated in the self-host build - Organization Creators, the Instance Management API, and UI customization. Everything you need - tracing, evals, prompt management, human annotation, RBAC - is free. It is framework-agnostic, runs as an OpenTelemetry backend, and had over 2,000 paying customers at acquisition, so this is a proven, high-install-base stack, not a science project.
The catch is operational. v3 moved from “just Postgres” to Postgres plus ClickHouse plus Redis plus S3-compatible storage - four services to stand up and keep running - and the migration to that architecture is where self-hosters get stuck, with GitHub threads reporting memory errors and timeouts. gRPC OTLP is not supported yet, HTTP only. And it is a ClickHouse subsidiary after the January 2026 acquisition, worth weighing on a multi-year bet. If you have a team that can run a four-service stack, this is the safest self-host in the category.
2. Opik - the cleanest license, zero feature gates
Opik edges Langfuse on the license and loses only on track record. Comet’s platform is Apache-2.0 with the full feature set self-hosted - unlimited spans, members and retention, and no feature gates at all. Where Langfuse gates three features, Opik gates none, and it is explicitly designed to self-host at Kubernetes scale without losing features. It captures multi-step traces, ships LLM-judge and code metrics, Guardrails and an Agent Optimizer, and it is the fastest-growing project of its peers at roughly 20.8k GitHub stars. Backed by Comet, an established company since 2017 with roughly $63M raised, so the maturity is real.
The gotcha only bites on the cloud, not the self-host. Per-seat pricing scales poorly as your team grows on the managed tiers, and a few users report UI slowdown on very large projects. Neither touches the Apache-2.0 self-host, which stays free and full-featured. If the cleanest license with zero gates is your priority, Opik is the pick - Langfuse is only ahead on sheer install base.
3. Laminar - the only one you self-host end to end, built for agents
Laminar has the strongest self-host story after Langfuse, and a genuinely unique claim - it is the only platform here you can self-host in full. Not a gateway, not a hobbled core - the whole stack is open source and documented for self-hosting. It is OpenTelemetry-native, written in Rust for low overhead, and sharpest on agents, especially browser agents - Browser Use documents it as its observability integration, and OTel co-creator Ben Sigelman is an angel investor. If your workload is agents and data residency is non-negotiable, this is the targeted pick.
Two honest reservations. It is the youngest and smallest here (founded 2024, YC S24, $3M seed), which is a real risk for a conservative multi-year bet - though full open-source self-host gives you an exit if the company changes course. And the exact feature delta between self-host and cloud is not clearly documented, so confirm what you keep and what you lose before you build a production stack on the self-hosted edition. Managed billing (data GB plus the hard-to-forecast Signals meter) disappears when you self-host, which is a point in its favor.
4. Arize Phoenix - free self-host, but read the license
Arize Phoenix self-hosts free with no usage caps and no feature gates on the features themselves, and it is famously fast - a working trace UI on your laptop in under a minute. It is OpenTelemetry-native with strong eval, including the best RAG evaluation in the category. For internal self-hosted use it is excellent.
The license is why it sits last in a self-host roundup specifically. Arize markets Phoenix as “fully open source, no feature gates,” but the main server repo is Elastic License 2.0 - source-available, not OSI-approved open source. Only the client and eval subpackages are Apache-2.0. ELv2 forbids offering Phoenix as a hosted service to third parties. For a team self-hosting internally that restriction never comes up, and the features are not gated - the license is. But in a roundup ranked on license honesty, ELv2 is a narrower freedom than MIT or Apache-2.0, and that is the tiebreaker. There are also reports of ingest lag before traces appear.
So which one?
- You want the battle-tested self-host default and can run four services - Langfuse.
- You want the cleanest license with zero feature gates - Opik, Apache-2.0 and full-featured self-hosted.
- You are self-hosting an agent stack end to end - Laminar, confirming the self-host feature delta first.
- You want fast, capable self-host for internal use - Arize Phoenix, as long as you are not reselling it as a service.
Every license, price and date above is read from each vendor’s own materials, verified on 23 July 2026, and this category ships breaking changes monthly - we re-verify every 30 days. The full breakdowns live in our tool reviews. For self-host, the honest split is Langfuse for maturity, Opik for the cleanest license - and both give you the real product for free.
Frequently Asked Questions
What is the best self-hosted LLM observability tool in 2026?
Langfuse. It is MIT-licensed, self-hosts free, and only three features are enterprise-gated, so tracing, evals, prompt management and annotation are all included. It is the most battle-tested self-host in the category. The catch is the ops burden - v3 needs Postgres plus ClickHouse, Redis and S3-compatible storage, four services, and the migration is where self-hosters get stuck. Opik is the close alternative if you want zero feature gates.
Which self-hosted tool has the cleanest license?
Opik. Its Apache-2.0 build ships the full feature set self-hosted with unlimited spans, members and retention and no gates at all. Langfuse is MIT with only three features gated. Laminar is fully open-source and the only one you can self-host end to end. Phoenix is the asterisk - its server is Elastic License 2.0, source-available rather than OSI open source, which forbids offering it as a hosted service to third parties.
What is the real cost of self-hosting LLM observability?
Operations, not licensing. The software is free on all four, but you run the infrastructure. Langfuse v3 is a four-service stack (Postgres, ClickHouse, Redis, S3) and the migration bites. Opik self-hosts at Kubernetes scale. Laminar runs the whole stack yourself in Rust. Phoenix famously starts locally in under a minute. Budget on-call and infra time, not license fees.
Can I self-host to control data residency?
Yes, that is the main reason to self-host - your trace data never leaves your infrastructure. All four here support it. Langfuse (MIT), Opik (Apache-2.0) and Laminar (fully open-source) let you run the whole thing free. Phoenix OSS self-hosts free too, with the Elastic License 2.0 caveat only mattering if you plan to resell it as a service.
Explore More
Tool Reviews
Related Articles
- How to Benchmark AI Agents in 2026 - The Tools and the Method
- The Best LLM Tracing Tools in 2026, Ranked by OpenTelemetry Depth
- The Best Open-Source LLM Observability Tools in 2026, Ranked by License Reality
- The Best OpenTelemetry LLM Observability Tools in 2026, Ranked
- The Best LLM Observability Tools in 2026, Ranked and Road-Tested
Free Newsletter
Get the LLM Evals Newsletter
Platform comparisons, pricing changes and eval technique deep-dives. No spam.
Related Articles
How to Benchmark AI Agents in 2026 - The Tools and the Method
Benchmarking an agent is not benchmarking a model. Public leaderboards tell you about the LLM, not your agent on your task. Here is how to build a real agent benchmark, and the five tools that actually run one - simulation, datasets, trajectory scoring and repeatable eval sets, ranked.
July 26, 2026
best-ofThe Best AI Agent Observability Tools in 2026, Ranked for Multi-Step and Browser Agents
Four platforms for tracing agents that loop, call tools, and click around browsers - judged on agent-native tracing, self-host reality, pricing you can forecast, and pre-release testing. One purpose-built winner, and where each meter bites.
July 26, 2026
best-ofThe Cheapest LLM Observability Tools in 2026, Ranked by Real Cost
The three lowest-cost ways to get production LLM tracing - the cheapest managed cloud, the cheapest self-host, and the free tier that looks great until you read the fine print. Priced at the tiers you will actually hit.
July 26, 2026
Langfuse Review
Opik Review
Laminar Review
Arize Phoenix Review