The 2026 LLM Observability Consolidation Map - Who Got Bought, Who Stayed Free

Early 2026 was the season the LLM observability market started to consolidate. Three tools that a lot of teams depend on changed hands in under two months. One got a stable new parent. One got frozen. One got folded into the biggest lab in the industry.

If you are choosing an observability platform right now, ownership is no longer a footnote. The difference between a tool that gets invested in and a tool that gets sunset under you is the difference between a two-year headache and a clean stack. So here is the map - the three acquisitions, what each one actually means for a buyer, who is still independent, and how to choose so you do not get stranded.

Every date, valuation and funding figure below is drawn from our individual tool reviews and links back to them.

Langfuse to ClickHouse - the stable one

ClickHouse acquired Langfuse on 16 January 2026, announced alongside ClickHouse’s own $400M Series D at a $15B valuation. At the time Langfuse had over 2,000 paying customers and was used by 19 of the Fortune 50. It was founded in 2022 in Berlin and came through Y Combinator’s W23 batch.

The reassuring part is what did not change. ClickHouse committed to keeping the MIT license, free self-hosting as a first-class option, and the roadmap intact, and so far that has held. Langfuse already ran its heavy storage on ClickHouse, so this is a natural fit rather than an awkward bolt-on. The self-host build stays near-complete - only three features are enterprise-gated - and the economics that make it the category’s open-source default are unchanged, still roughly $101/mo at 1M events against LangSmith’s ~$2,514/mo.

What it means for buyers. This is the safest of the three outcomes. You are not betting on a struggling startup - you are betting on a subsidiary of a well-funded database company that has every reason to keep Langfuse strong, because it drives ClickHouse adoption. The only thing worth filing away is the strategic one - if ClickHouse’s priorities ever shift, a database company’s incentives are not identical to an observability company’s. For now, though, Langfuse remains the tool I recommend most, acquisition and all.

Helicone to Mintlify - the frozen one

This is the cautionary tale. Mintlify acquired Helicone on 3 March 2026, and Helicone is now in maintenance mode. Only security patches, bug fixes and new-model support ship. No new integrations, no new analytics, no roadmap. And Mintlify has said it is actively helping existing customers migrate off. At acquisition, Helicone had processed 14.2 trillion tokens across roughly 16,000 organizations, so this was not a fringe tool.

Helicone was a clean, open-source, proxy-based observability layer under Apache 2.0. In a different timeline it would still be an easy recommendation. But the maintenance-mode status overrides everything else. Self-hosting the Apache-licensed code still technically works, but you would be maintaining a frozen codebase whose own creator is pointing users to the door - a much weaker bet than a self-host of a tool that keeps improving.

What it means for buyers. Do not start a new project on Helicone. Full stop. It does not matter how tidy the product is when the people who own it are winding it down. If you are already on Helicone, you are not in immediate danger - security fixes still ship - but you should treat the migration push as a countdown and plan your exit deliberately rather than under pressure later. The natural landing spots are Langfuse or Opik for open-source observability, or Portkey if it was the gateway angle you valued.

Promptfoo to OpenAI - the strategic one

OpenAI acquired Promptfoo on 9 March 2026, six days after the Helicone deal. Promptfoo is the de-facto open-source CLI for LLM eval and red-teaming - MIT-licensed, driven by declarative YAML, with roughly 23.5k GitHub stars, the most of the major eval tools. Before the acquisition it had raised about $23.4M - a $5M seed from a16z, then an $18.4M Series A led by Insight Partners in July 2025.

The commitment is on the record. Both OpenAI and Promptfoo state the project stays open source under its current MIT license, existing customers keep being supported, and the technology folds into OpenAI’s “Frontier” work on agentic security testing. So unlike Helicone, this is not a wind-down - the OSS continues, and for security and CI teams the free tier is still the real product.

What it means for buyers. The license protects you, which is the whole point of MIT - even if OpenAI’s priorities change, the existing open-source code cannot be clawed back. The open question is governance and direction. Strategic control of the project now sits with OpenAI, and long-term stewardship of an OSS tool under a single large owner is a fair thing to weigh, especially if you care about vendor neutrality in your eval tooling. For most teams it is a fact to note, not a reason to walk away. If independence is a hard requirement, that is a point in favor of a still-independent eval tool.

Who is still independent

The good news for buyers worried about consolidation - most of the field has not moved. Here is where the still-independent players stand, by funding and valuation.

• LangSmith (LangChain, Inc.) - roughly $260M raised total, at a $1.25B valuation after a $125M Series B led by IVP in October 2025, following a $25M Series A led by Sequoia in early 2024. Well-funded and independent - survival is not the question here, only the price and the lock-in.
• Braintrust - raised an $80M Series B led by Iconiq in February 2026 at a roughly $800M valuation, on top of a $36M Series A led by a16z in October 2024. Growing fast and funded to stay.
• Arize - the company behind Phoenix raised a $70M Series C in February 2025 led by Adams Street Partners, with M12, Datadog and PagerDuty among investors. Founded 2020 in Berkeley.
• Opik (Comet) - Comet, founded 2017, raised a $50M Series B in November 2021 led by OpenView, roughly $63M total, and launched Opik under Apache-2.0 in September 2024. A mature company’s open-source play.
• Galileo - the best-funded platform in the space at roughly $68M total, including a $45M Series B in October 2024 led by Scale Venture Partners, with strategic money from Databricks Ventures and ServiceNow Ventures.
• Maxim - the youngest-funded of the serious players, a $3M seed in June 2024 led by Elevation Capital. Real product, small balance sheet.
• Portkey - raised a $3M seed then a $15M Series A led by Elevation Capital with Lightspeed. Founded January 2023.
• Laminar - the newest, founded 2024, a $3M seed, YC S24, with OTel co-creator Ben Sigelman as an angel investor. Fully open-source, so you have an exit even at its size.

What this means if you are choosing today

The consolidation does not mean “avoid tools that might get acquired” - almost any of these could. Langfuse got acquired and is still my top pick. The point is subtler. What protects you is not the tool staying independent forever. It is whether the outcome, if it gets bought, leaves you stranded.

Two things decide that.

First, the license. A permissive OSS license - MIT for Langfuse and Promptfoo, Apache-2.0 for Opik and the Helicone code - means the existing product cannot be taken away from you, whatever happens to the company. That is why Promptfoo joining OpenAI is a governance question, not an emergency. It is also why even Helicone’s frozen state is survivable for existing self-hosters, if a dead end. A closed-source tool gives you no such floor - if LangSmith, Galileo or Maxim ever changed course, you would have no code to fall back on.

Second, settled ownership. A tool whose ownership question is resolved - either genuinely independent and well-funded, or acquired by a parent with aligned incentives - is a safer multi-year bet than one that looks like an acquisition target with a closed core. Langfuse under ClickHouse is settled. LangChain, Braintrust, Arize, Comet and Galileo are independent and funded.

The real lesson of early 2026 is Helicone. A clean, capable, open-source tool got absorbed and put into maintenance mode within weeks, and its own new owner started helping customers leave. That is the failure mode to design against. Favor tools whose ownership is settled, or whose OSS license means a bad outcome still leaves you with a product you control. On both tests, the open-source platforms - Langfuse, Opik, Laminar - come out ahead, which is a large part of why they top our observability roundup.

Every date, valuation and funding figure on this page is drawn from our individual tool reviews and was verified against each vendor’s materials on 23 July 2026. This category moves fast - we re-verify every 30 days, and we will update this map as the next deals land.