Overview
Helicone began as a one-line proxy for logging OpenAI requests and grew into an open-source AI gateway and observability suite. Its product followed the operational needs of AI teams: first cost and request visibility, then routing, fallbacks, prompt management, evaluations, and release workflows. By March 2026, the company reported processing 14.2 trillion tokens for 16,000 organizations and tracking 33 million end users.[1]
Mintlify acquired Helicone on March 3, 2026. The outcome was not a shutdown, but it was not an independent growth ending either: the founders joined Mintlify, Helicone entered maintenance mode, and customers were offered support migrating elsewhere.[2] The strongest reading is that Helicone won adoption while the standalone observability category lost strategic altitude. Basic tracing became available from open-source competitors and model providers; the more valuable prize moved toward the knowledge infrastructure that agents use to act.
Founding Story
Helicone came out of Y Combinator's Winter 2023 batch in San Francisco. YC identifies Justin Torre, Barak Oshri, and Scott Nguyen as founders. Torre had worked in developer evangelism and teaching at Apple, Oshri had been a machine-learning engineer at Sisu Data and worked in Stanford's AI Lab, and Nguyen brought user-experience and finance experience.[3] Cole Gottdank later appeared as a founder and co-signatory of the acquisition announcement. Public sources reviewed here do not explain the founder lineup change or how the original team met.
The founding wedge was deliberately small. Developers changed a base URL, routed model calls through Helicone, and received logs plus cost, user, model, and prompt analytics. Caching and intelligent retries addressed reliability and spend without forcing a team to replace its application stack.[3] That low-friction entry mattered because AI application teams in 2023 were shipping before a stable tooling category existed.
The founders later wrote that YC batchmates asked for access, continued using the product, and referred friends. They described Helicone as an early mover before “LLM observability” had settled into a recognized market.[1] The evidence handoff does not contain two complete, attributable founder quotations, so this report does not manufacture them. That source gap limits what can be said about the founders' private motivations and internal decisions.
Timeline
- Winter 2023: Helicone joined Y Combinator and launched around proxy-based LLM logging, analytics, caching, and retries.[3]
- January 23, 2025: Helicone launched V2 on Hacker News as a workflow spanning “Log → Evaluate → Experiment → Review → Release.” It reported 2.1 billion requests and 2.6 trillion tokens processed.[4]
- March 3, 2026: Mintlify announced its acquisition of Helicone. Justin Torre and Cole Gottdank joined Mintlify; Helicone moved into maintenance mode.[1]
What They Built
Helicone sat between an AI application and model providers. A developer could route a request through its gateway and gain a record of what was sent, what came back, latency, cost, user context, and errors. That architecture made instrumentation an infrastructure choice rather than a recurring application-code project.
The product expanded in both directions. At the traffic layer it added routing, automatic provider fallback, retries, caching, and cost controls. At the development layer it added prompt versioning, evaluations, experiments, agent-session debugging, and review-to-release workflows. Its current open-source repository describes a Next.js web app, a Cloudflare Workers proxy, an Express/Tsoa collector, Supabase for database and authentication, ClickHouse for analytics, and object storage through Minio.[5]
Scale forced architectural change. Helicone said synchronous log processing overloaded its database and pushed queries from milliseconds to minutes. The team rebuilt ingestion around Kafka and split storage by access pattern across Kafka, S3, and ClickHouse. It reported zero data loss and fast queries at billions of records after the change.[4] This was more than backend housekeeping: a monitoring product that drops events or stalls under load destroys the trust it sells.
The distinguishing package was breadth with deployment control. Teams could use a hosted gateway, bring provider keys, or self-host under Apache 2.0. Competitors often owned one layer well, such as tracing, prompt evaluation, or routing. Helicone tried to join those layers around the request path. That integration increased utility, but also placed the company between open standards below and larger application platforms above.
Market Position
Target Customers
Helicone initially fit small AI application teams that needed visibility without a dedicated platform group. The one-line proxy reduced adoption cost, while open source and self-hosting widened the audience to security-conscious enterprises. As the product grew, it also addressed platform engineers managing several model providers, finance teams allocating inference spend, and developers debugging multi-step agents.
Market Size
No reliable primary source reviewed here provides a defensible market-size estimate. Helicone's self-reported 16,000 organizations, 14.2 trillion processed tokens, and 33 million tracked end users show broad usage, not revenue or total market value.[1] The absence matters. Large token throughput can coexist with weak monetization when free tiers, bring-your-own-key traffic, and zero-markup routing dominate.
Competition
Langfuse covers open-source tracing, prompt management, evaluation, analytics, and playground workflows, and reports more than 2,300 companies and billions of observations per month.[6] Traceloop's OpenLLMetry takes a standards-first route: Apache-2.0 OpenTelemetry instrumentation exportable to more than 25 observability backends.[7] OpenAI's Agents SDK includes native tracing for generations, tool calls, handoffs, guardrails, and custom events.[8]
Those rivals squeezed Helicone from three sides. Open-source peers reduced switching costs, generic observability vendors benefited from existing enterprise contracts, and providers could bundle tracing close to their models. Helicone's gateway created a stronger control point through routing and failover, but its published model of 0% markup and $20 per seat constrained direct participation in model spend.[9]
Humanloop's platform sunset in September 2025 provides another category signal.[10] It does not prove that observability lacked demand. It suggests that broad AI-development suites faced consolidation pressure as primitives became features and strategic buyers pursued teams, technology, and adjacent infrastructure.
Business Model
Helicone offered a hosted product, self-hosting, and a free tier of 10,000 requests per month. Its documentation described pass-through provider billing with no gateway markup and a $20-per-seat comparison point.[5][9] The model reduced resistance for developers but detached revenue from the most visible growth metric: token volume.
No primary source reviewed here disclosed revenue, annual recurring revenue, gross margin, burn, retention, or the paid/free customer mix. Funding beyond verified YC participation is also unconfirmed. Purchase price, consideration, earn-outs, investor returns, and liabilities in the Mintlify deal were not disclosed by either party. Any claim that the acquisition was financially successful or distressed would therefore be speculation.
Traction
Helicone reported 2.1 billion requests and 2.6 trillion tokens processed by January 2025, serving customers from startups through Fortune 500 companies.[4] By the acquisition announcement, it claimed 14.2 trillion tokens, 16,000 organizations, and 33 million end users tracked.[1]
Distribution came through YC referrals, open-source GitHub discovery, the low-effort proxy integration, Discord, self-hosting, and Product Hunt. The founders claimed a number-one Product Hunt launch and described Helicone as the most-used LLM-observability platform among YC companies.[1] These figures and rankings are self-reported and were not independently verified.
Post-Mortem
Adoption did not guarantee standalone economics
Helicone's usage curve is the strongest counterargument to any product-market-fit failure story. Customers routed enormous traffic through the product, and the team solved difficult scale problems. Yet the commercial evidence required to connect usage to a durable independent company is absent. A free tier, bring-your-own keys, and 0% gateway markup made adoption easier while limiting how directly Helicone captured the value of model traffic.
The response was product expansion. V2 moved from logs toward evaluations, experiments, reviews, and releases.[4] That increased surface area just as competitors and providers bundled overlapping functions. The mechanism was not simply competition. Helicone's low-friction wedge attracted traffic faster than it established a proprietary system of record, while broadening the suite raised maintenance demands.
The category's primitives became portable
OpenTelemetry-based instrumentation and provider-native tracing weakened observability as a proprietary layer. Teams could export traces to existing backends or accept tools bundled by model vendors. Helicone's gateway remained strategically useful because it could route across providers, but its observability interface faced declining differentiation.
Mintlify's acquisition supports this interpretation. Its announcement tied the team to intelligent routing, real-time observability, model failover, and AI knowledge infrastructure, while Helicone itself entered maintenance mode.[2] The acquired capability mattered inside a broader platform more than the standalone product apparently did. That is an inference from the announced operating plan, not a disclosed rationale or valuation judgment.
Maintenance mode converts platform risk into customer risk
Open source and self-hosting protected customers from total lock-in, but a maintenance-mode acquisition still changes the trust equation. Security fixes and new-model support may continue, yet roadmap ambition and hosted-service continuity become uncertain. Mintlify said customers would receive migration support, a candid signal that continued standalone use was not the only expected path.[2]
The acquisition letter made the product's reduced ambition explicit: “Helicone's services will remain live for the foreseeable future in maintenance mode.”[1] Maintenance preserves service, but it does not promise the active roadmap customers had originally bought into.
The structural lesson is specific: an observability gateway asks customers to place sensitive traffic and operational history in its path, but an acquisition can strand the interface and data model even when the code remains available. Durable export formats, policy portability, and independently operable data planes should be product features, not exit-plan paperwork.
Key Lessons
- A one-line wedge can win the market before it wins the business. Helicone reduced integration friction and earned substantial self-reported usage. Its undisclosed revenue and maintenance-mode outcome show why traffic needs a clear capture mechanism.
- Reliability is part of the monitoring product. The Kafka, S3, and ClickHouse rebuild answered database overload with architecture matched to access patterns. Losing or delaying telemetry would have negated the product's promise.
- Broad suites collide with bundling. Helicone expanded from logging into evaluation and release workflows while open-source and provider-native tools covered the same ground. A standalone vendor needs a control point incumbents cannot cheaply absorb.
- Portability must survive the vendor. Apache-2.0 licensing and self-hosting helped, but maintenance mode still created migration risk. Open trace, policy, and audit formats would let customers retain operational continuity when ownership changes.
Sources
- Helicone, “Joining Mintlify”
- Mintlify, “Mintlify acquires Helicone”
- Y Combinator company profile: Helicone
- Hacker News: Helicone V2 launch
- Helicone GitHub repository
- Langfuse
- OpenLLMetry GitHub repository
- OpenAI Agents SDK tracing documentation
- Helicone platform overview
- Humanloop platform sunset notice