Overview
Traceloop built observability and evaluation infrastructure for teams putting large-language-model applications into production. Founded in 2023 and admitted to Y Combinator's Winter 2023 batch, the nine-person Tel Aviv company paired OpenLLMetry, its OpenTelemetry-based instrumentation project, with a commercial platform for tracing, quality checks, prompt experiments, and drift monitoring.[1]
Traceloop's outcome was an acquisition, not a shutdown. Open source gave the company reach across an unusually fragmented AI stack; the paid layer turned that telemetry into enterprise controls. ServiceNow bought the missing runtime-evidence layer for AI Control Tower, completing the acquisition by May 5, 2026.[2] The consideration, deal structure, precise closing date, and employee transition remain undisclosed.
Founding Story
Nir Gazit and Gal Kleinman came from the same professional orbit: both had worked at Fiverr, where Gazit became chief architect and Kleinman managed an R&D group. Gazit had also worked at Google.[1] Gazit's later account dates the work to 2022, when the pair were watching GPT-3 applications behave unpredictably while teams tracked prompt changes in spreadsheets.[3] The specific circumstances in which the founders met are not disclosed in the observed material beyond their shared Fiverr background.
Their initial bet was infrastructure before interface. OpenLLMetry extended OpenTelemetry so developers could instrument calls across model providers, agent frameworks, and vector databases, then export traces through standard OTLP plumbing rather than accept a proprietary sink.[4] That architecture was also a distribution decision. A developer could add tracing with a small SDK change and continue using destinations such as Datadog, Honeycomb, Sentry, Grafana, or ServiceNow Cloud Observability. The commercial product then moved up the stack, adding evaluation, prompt versioning, model comparison, and cost and performance monitoring.[3]
The evidence handoff contains founder-authored accounts but no verbatim passages observed from the full pages. This report therefore does not manufacture the two direct founder quotations requested by the canonical format. That gap matters less to the chronology than to motive: Gazit's retrospective supports the spreadsheet-and-unpredictability origin, but no independently observed founder interview tests how the founders chose observability over adjacent gateway or development-tool wedges.
Timeline
- 2022: Gazit says he and Kleinman began work after encountering unpredictable GPT-3 systems and spreadsheet-based prompt management.[3]
- 2023: Traceloop was founded and joined YC's Winter 2023 batch.[1]
- December 2024: Support documentation recorded a $59 monthly Growth tier for the first 100,000 spans, plus usage charges.[5]
- May 2025: Gazit announced a $6.1 million seed round and reported OpenLLMetry above 500,000 monthly downloads.[3]
- March 2026: Traceloop announced that it was joining ServiceNow and that its technology would enter AI Control Tower.[6]
- May 5, 2026: ServiceNow said the acquisition had been completed.[2]
What They Built
OpenLLMetry was the collection layer. It added vendor-neutral semantic conventions and instrumentation around model calls, agent steps, retrieval operations, and related application work. Its public repository documents support across OpenAI, Anthropic, Bedrock, Gemini, Cohere, Groq, Mistral, and Vertex; frameworks including LangChain, LlamaIndex, CrewAI, LangGraph, LiteLLM, and MCP; and vector databases such as Pinecone, Qdrant, and Weaviate.[4] When observed, the repository showed about 7,100 stars, 978 forks, and 1,395 commits. Those are repository snapshots, not customer or revenue measures.
The user journey began with instrumentation. A developer initialized the SDK, emitted telemetry through OpenTelemetry-compatible infrastructure, and inspected the resulting traces. The paid platform added operational interpretation: faithfulness and relevance checks, custom evaluators, drift detection, prompt versioning and testing, model comparisons, and cost and latency monitoring.[3] This separated portable evidence collection from proprietary analysis.
That distinction proved strategically useful. Dynatrace published an OpenLLMetry integration guide, while IBM documented OpenLLMetry in its GenAI observability work.[7][8] These pages verify ecosystem integration, not paid Traceloop contracts. Gazit also named Cisco and IBM as users of the open-source project and Miro as a commercial customer experimenting at scale, but comparable first-party Cisco or Miro case studies were not observed.[3]
The product did not disappear after the transaction announcement. Traceloop's site still presented sign-in, documentation, status, pricing, and on-premises paths, and its pricing page listed a free tier with 50,000 spans per month and five seats plus a sales-led enterprise tier.[9] Gazit's announcement said OpenLLMetry would remain open source and existing obligations would be honored.[6] ServiceNow's longer-term packaging and roadmap were not disclosed.
Market Position
Target Customers
Traceloop straddled two buyers. OpenLLMetry addressed developers who needed portable tracing across a changing set of models and frameworks. The commercial platform addressed AI platform and reliability teams that needed evaluations, drift alerts, retention controls, security assurances, and deployment options. The current free tier supports developer adoption; enterprise features such as SOC 2 support, custom retention, and on-premises or air-gapped deployment point toward regulated or security-sensitive organizations.[9]
Market Size
No observed source supplies a defensible total-addressable-market estimate for Traceloop's exact category. The better evidence is strategic demand. ServiceNow described AI Control Tower as a system for discovering, observing, governing, securing, and measuring AI across enterprise systems, and said Traceloop added visibility into agent reasoning, decisions, and course correction.[2] The acquisition validates runtime evidence as part of enterprise AI governance, but it does not disclose market share, revenue, or growth.
Competition
The competitive axes were portability, workflow breadth, deployment control, and distribution. Langfuse offered open-source tracing, evaluations, prompts, datasets, experiments, cloud hosting, and self-hosting.[10][11] Helicone paired observability with an AI gateway, routing, caching, rate limits, and prompt management, with public plans ranging from free to enterprise and on-premises deployment.[12][13] Humanloop leaned toward enterprise evaluation workflows, including offline and log-based evaluation and self-hosting.[14]
Provider-native tooling created the deeper structural pressure. OpenAI's Agents SDK included tracing, while AgentKit added datasets, trace grading, prompt optimization, and third-party model support.[15][16] A standalone vendor could not defend generic trace viewing forever. Traceloop's strongest position was therefore not a prettier dashboard. It was neutral instrumentation plus enterprise evaluation and governance across providers. ServiceNow could distribute that capability inside a broader control plane rather than sell observability as an isolated tool.
Business Model
Traceloop used open source and a free hosted tier to seed adoption, then charged for greater volume and enterprise controls. A December 2024 support article recorded a $59 monthly Growth plan covering the first 100,000 spans, $10 per additional 10,000, and a separate $0.01 monitored-span charge.[5] Its current public page instead exposes Free Forever and a sales-led Enterprise tier.[9] AWS Marketplace listed Enterprise Edition at $15,000 for 12 months, with larger discounts for longer terms; that marketplace listing should not be treated as the current direct contract price.[17]
This model aligned payment with telemetry volume, retention, evaluations, support, and deployment requirements. It also exposed a tension: open standards made adoption easier but let users send data elsewhere. The commercial product had to earn its margin through evaluation workflows and governance, not ingestion alone. Traceloop disclosed neither revenue nor customer concentration. With no acquisition price, valuation, or consideration available, investor returns and transaction economics cannot be estimated responsibly.
Traction
Gazit reported more than 500,000 monthly OpenLLMetry downloads in May 2025 and named Cisco, Dynatrace, and IBM as users; he also named Miro as a customer using the platform at scale.[3] The repository snapshot of roughly 7,100 stars and 978 forks offers a separate signal of developer interest.[4] Dynatrace and IBM documentation independently confirms ecosystem use of OpenLLMetry.[7][8]
These measures establish distribution, not commercial scale. Monthly package downloads can include automation and repeated installs. GitHub stars do not equal production deployments. No observed source disclosed annual recurring revenue, paid-customer count, renewal rates, gross margin, evaluation cost, or enterprise pipeline. Traceloop's nine-person YC listing and $6.1 million seed indicate a compact, venture-backed team, but they do not support a burn estimate without salaries, infrastructure spend, or period data.[1][3]
Post-Mortem
The open standard became the acquisition surface
There is no failure cause to diagnose. The central mechanism is that Traceloop separated a neutral collection standard from its paid interpretation layer. OpenLLMetry could spread because it worked with existing observability destinations and a broad set of providers and frameworks.[4] That portability constrained lock-in revenue, yet it also made the technology useful to an enterprise platform that needed evidence across heterogeneous AI systems.
ServiceNow did not describe the asset as a standalone dashboard. It placed Traceloop inside AI Control Tower, where runtime traces inform governance, security, measurement, and intervention.[2] The non-obvious lesson is that limited proprietary control at the instrumentation layer can increase strategic value when the acquirer owns the system of action above it.
Generic observability was being compressed
The counter-narrative is that the acquisition alone proves an independent category winner. Competition argues for more caution. Langfuse and Helicone covered much of the open-source tracing and evaluation surface, Humanloop specialized in evaluations, and OpenAI moved tracing and grading into its own agent tooling.[10][12][14][16] As providers absorb baseline diagnostics, neutral vendors must win on cross-provider evidence, regulated deployment, and governance workflows.
Traceloop's response was visible before the sale: move from traces into custom evaluators, drift detection, prompt experiments, and enterprise deployment.[3] ServiceNow offered a broader control plane and enterprise distribution. The observed evidence cannot show whether Traceloop faced commercial pressure, chose an unusually attractive strategic exit, or both. Revenue, churn, runway, competing bids, and board deliberations are absent.
The terms limit any return analysis
Traceloop announced that it was joining ServiceNow in March 2026; ServiceNow confirmed completion by May 5.[6][2] Neither source disclosed price, cash-versus-stock structure, retention packages, exact closing date, or transferred headcount. The hosted product remained available and the founder committed to existing obligations, so descriptions such as shutdown, acqui-hire, or distressed wind-down would exceed the record.[6]
The funding record also contains a smaller unresolved inconsistency. Gazit's May 2025 post named Sorenson Ventures, YC, Samsung Next, IBEX, and angels in the $6.1 million seed.[3] The acquisition announcement described Ibex Investors, Sorenson Capital, and Grand Ventures as early-2025 backers.[6] This may reflect naming or round-description drift, but the observed sources do not resolve it.
Key Lessons
- Standards can be strategic assets. OpenLLMetry reduced lock-in by exporting through OpenTelemetry, but that neutrality built integrations and distribution. ServiceNow valued the runtime evidence because AI Control Tower had to govern systems it did not own end to end.
- The paid layer must sit above the commodity. Open tracing attracted developers; evaluations, drift detection, retention, and air-gapped deployment gave enterprises reasons to pay. As model providers added native traces and grading, generic dashboards lost room to differentiate.
- Distribution metrics need commercial counterparts. More than 500,000 monthly downloads and thousands of GitHub stars document reach, not revenue quality. Without paid-customer, retention, or margin data, observers cannot turn community adoption into a valuation claim.
- An acquisition announcement is not an exit return. Completion is confirmed, but price, structure, headcount transfer, and investor proceeds are not. The responsible conclusion is strategic fit, not financial victory or distress.
Sources
- Y Combinator: Traceloop
- ServiceNow: AI Control Tower expansion
- Traceloop: From vibes to visibility
- GitHub: OpenLLMetry
- Traceloop support: pricing and span management
- Traceloop: joining ServiceNow
- Dynatrace: OpenLLMetry integration
- IBM: GenAI and LLM observability
- Traceloop pricing
- Langfuse pricing
- Langfuse self-hosting
- Helicone platform overview
- Helicone pricing
- Humanloop evaluators
- OpenAI Agents SDK tracing
- OpenAI AgentKit
- AWS Marketplace: Traceloop Enterprise Edition