Portkey
Open-source AI gateway and unified control plane for production AI — LLM routing, governance, observability, reliability and cost management for enterprise AI agents; founded 2023 by Rohit Agarwal and Ayush Garg, $15M Series A in 2024 led by Elevation Capital with Lightspeed participation, processing 1T+ tokens / day across 24,000+ organisations, and acquired by Palo Alto Networks in May 2026 as the AI Gateway anchor of the Prisma AIRS security platform.
The Business
Portkey is an open-source AI Gateway and unified control plane for production AI systems — LLM routing, governance, observability, reliability and cost management plus the MCP Gateway for AI-agent governance across enterprise tools and systems. The company was founded in 2023 by Rohit Agarwal (previously a founder of Framebench (collaborative video / creative review tool) acquired by Freshworks) and Ayush Garg, headquartered in Bengaluru with US distribution presence. Portkey’s open-source Gateway, fully open-sourced in March 2026, processes 1T+ tokens / day across 24,000+ organisations including Fortune 500 customers in finance, pharma and technology, with $180M+ in annualised AI spend managed through the gateway. Cumulative external venture capital is approximately $18M through the 2024 $15M Series A led by Elevation Capital with Lightspeed participation, plus a prior $3M seed in August 2023 led by Lightspeed. On April 30, 2026, Palo Alto Networks (NASDAQ: PANW) announced its intent to acquire Portkey to anchor the AI Gateway core of the Prisma AIRS comprehensive AI security platform; the acquisition was completed May 29, 2026.
Customers and Distribution
Portkey’s customer base is approximately 24,000 organisations as of the March 2026 open-source-Gateway disclosure, including Fortune 500 customers in finance, pharma and technology. Distribution sits across three principal motions: the open-source Portkey Gateway as the bottom-up developer-and-enterprise adoption channel (fully open-sourced March 2026 with 1T+ tokens / day processed and 120M+ AI requests / day), the managed Portkey Gateway with advanced governance, observability and compliance features for enterprise customers, and post the May 29, 2026 Palo Alto Networks acquisition closing the Prisma AIRS comprehensive AI security platform enterprise channel through Palo Alto Networks’ existing enterprise-security customer base. Specific commercial revenue figures and named customer disclosures at filing-grade granularity are not separately broken out in primary sources; we anchor to the 1T+ tokens / day usage and 24,000+ organisation install base as the principal commercial-scale indicators and reference the post-acquisition Palo Alto Networks segment-reporting cycle as the future filing-grade source.
Model Strategy
Portkey is a Verticals-first generative-AI play under the IM Framework eight-trajectories taxonomy as it applies to AI infrastructure: the strategic bet is that depth on the AI Gateway primitive — LLM routing, governance, observability, reliability, cost management and MCP-based AI-agent governance — for production AI workloads is itself a defensible category that the foundation-model-provider native control planes and the hyperscaler-cloud-native AI Gateway alternatives do not fully match. The architecture is structurally multi-supplier and multi-model — Portkey routes across OpenAI, Anthropic, Google Gemini, open-weight model alternatives and hyperscaler-cloud AI Gateway integrations — with the D4a supplier-diversity sub-rubric held at 8 in the v1.6 evidence pass on that basis. Above the routing-and-observability layer, the differentiating product surfaces are the governance and policy framework, the reliability and cost-management features, the open-source distribution flywheel anchoring community adoption, and post-acquisition the Prisma AIRS security integration as the structural commercial enterprise channel. The portability profile is the defining structural variable — multi-cloud and multi-model architecture is the structural differentiator against the hyperscaler-cloud-native AI Gateway alternatives but foundation-model-provider native AI Gateway substitution and the post-acquisition integration execution inside Palo Alto Networks are the principal structural risks.
At A Glance
The Numbers
Trend charts are not shown for Portkey — only single-point data is currently available. See At A Glance above for the most recent disclosed values.
Leadership Team
Portkey is founder-led with Rohit Agarwal and Ayush Garg as the principal public voices. The executive layer below the co-founders has been built around early-stage engineering and developer-relations hires from the Indian and global AI-developer-tooling ecosystem (Bengaluru-anchored hub with US distribution presence). Post the May 29, 2026 completion of the Palo Alto Networks acquisition, Portkey integrates into the Prisma AIRS comprehensive AI security platform with the founders continuing to lead the AI Gateway product line inside Palo Alto Networks per the acquisition framing.
IM Framework Scoring
IM’s structured assessment of Portkey’s competitive position. The summary below is the headline; expand “Show the full analyst-grade analysis” near the bottom for the per-dimension reasoning and evidence. Methodology →
Funding History
| Date | Round | Raised | Post-money | Lead investor(s) |
|---|---|---|---|---|
| May 2026 | Acquisition — Palo Alto Networks | Acquisition — terms not separately disclosed | Acquired by Palo Alto Networks (NASDAQ: PANW) | Palo Alto Networks |
| 2024 | Series A | $15M | — | Elevation Capital (with Lightspeed) |
| Aug 2023 | Seed | $3M | — | Lightspeed |
Cumulative external venture capital approximately $18M through the 2024 $15M Series A led by Elevation Capital with Lightspeed participation per Entrackr’s coverage and Portkey’s own Series A announcement. Prior August 2023 $3M seed led by Lightspeed per Ventureburn. The May 29, 2026 acquisition by Palo Alto Networks (announced April 30, 2026) per Palo Alto Networks’ press release with specific deal terms not separately disclosed in the press cycle.
Competitive Landscape
| Competitor | Positioning | Distribution edge | Threat profile |
|---|---|---|---|
| Cloudflare AI Gateway ((NYSE: NET)) |
Cloudflare’s edge-native AI Gateway, bundled with Workers AI and Vectorize; pitched as the cheapest gateway-and-cache layer because it rides Cloudflare’s existing edge POPs. | Sells into Cloudflare’s multi-million-account developer base and global enterprise security/CDN install base; AI Gateway is a free-tier feature on existing Cloudflare plans. | High — Cloudflare’s AI Gateway is the principal hyperscaler-native AI Gateway alternative integrated into Cloudflare Workers AI distribution; competes head-to-head on the LLM-routing, governance and observability control plane with Cloudflare’s edge-network distribution as the structural advantage. |
| LangSmith / LangChain | LangSmith is the observability and eval product for the LangChain agent framework — overlaps with Portkey on observability but is narrower on routing / governance / cost-control. | Reaches developers via the LangChain open-source framework’s pull (one of the most-installed AI dev packages); cross-sells LangSmith into the LangChain user base. | Medium-high — LangChain’s LangSmith observability product is the principal LLM-observability competitor with deep developer-distribution from the broader LangChain open-source agent-framework ecosystem; flanking risk on the observability sub-surface rather than head-to-head on the full AI Gateway control plane. |
| Helicone | Symmetric pure-play LLM observability and gateway competitor; OSS-first proxy plus a hosted observability dashboard, pitched on developer ergonomics. | Open-source GitHub distribution and Y Combinator alumni network; bottoms-up developer adoption on the same surface Portkey contests. | Medium — LLM observability and gateway startup competing on the developer-tools surface with comparable open-source distribution; symmetric pure-play rival on the AI Gateway category. |
| OpenAI / Anthropic native control planes | Foundation-model providers extending into the gateway layer: prompt caching, model routing, evals and usage governance bundled into the model-provider’s own product surface. | Reach customers directly through the model-provider relationship the gateway exists to abstract; structural pressure on standalone gateways as multi-model orchestration moves upstream. | Medium-high — foundation-model-provider native AI Gateway alternatives (OpenAI Operator-adjacent control planes, Anthropic agent-gateway features) that bundle gateway, observability and governance into the model-provider product surface; structural pressure on the standalone AI Gateway category as foundation-model providers extend into the gateway layer. |
| Hyperscaler-native AI Gateway (AWS, Azure, GCP) | AWS Bedrock, Azure AI Foundry and Google Vertex bundle routing, observability, guardrails and governance into the hyperscaler AI platform — the default gateway story for the enterprises already on those clouds. | Reaches enterprise via existing AWS / Azure / GCP commit-spend; no separate procurement, which is the structural pressure on the standalone-gateway category. | Medium-high — AWS Bedrock, Azure AI Foundry and Google Vertex AI all increasingly bundle gateway, observability and governance features into the hyperscaler-cloud AI infrastructure subscription; bundled enterprise procurement against the standalone AI Gateway category. |
Potential Risks
Foundation-model-provider native AI Gateway substitution
OpenAI Operator-adjacent control planes, Anthropic agent-gateway features and the hyperscaler-cloud AI infrastructure subscriptions (AWS Bedrock, Azure AI Foundry, Google Vertex AI) all increasingly bundle gateway, observability and governance features into the model-provider and hyperscaler product surfaces. The structural risk is that as foundation-model providers and hyperscalers extend into the gateway layer, the standalone AI Gateway category compresses. The bull case is that Portkey’s multi-supplier / multi-cloud positioning plus the Palo Alto Networks Prisma AIRS security integration creates a distinctive structural moat that the native gateways cannot match for security-anchored enterprise customers; the bear case is that bundled native gateways from foundation-model providers and hyperscalers compress the standalone gateway category over the multi-year horizon.
Integration-execution risk inside Palo Alto Networks Prisma AIRS
Post the May 29, 2026 acquisition closing, Portkey integrates into the Prisma AIRS comprehensive AI security platform stack as the AI Gateway anchor. The structural risk is that AI Gateway integration execution inside a large enterprise-security platform is harder than independent product-led-growth motion; the integration cadence, the Palo Alto Networks enterprise-procurement channel conversion economics from the existing Portkey open-source 24,000+ organisation base, and the engineering and product-velocity dynamics are the principal execution variables. The bull case is that Palo Alto Networks’ enterprise security distribution is the structural unlock for the AI Gateway category at scale; the bear case is that integration friction slows the AI Gateway commercial trajectory below the independent product-led growth path.
Symmetric competitor pressure — Cloudflare AI Gateway, LangSmith, Helicone
Portkey’s most direct competitive risk is from Cloudflare AI Gateway (hyperscaler-native AI Gateway alternative with edge-network distribution), LangSmith / LangChain (observability competitor with deep developer-distribution from the LangChain ecosystem), and Helicone (symmetric pure-play rival on the AI Gateway category). The structural risk is not that any one rival beats Portkey head-to-head — the open-source-Gateway distribution flywheel is durable — but that the symmetric-competitor cadence compresses pricing-power and slows the commercial-trajectory growth even inside Palo Alto Networks.
Open-source business-model conversion economics
Portkey’s March 2026 transition to fully open-source Gateway is a deliberate strategic feature — it anchors community adoption and drives the 24,000+ organisation install base — but the same feature caps the share of usage that converts into commercial AI Gateway revenue. Post the Palo Alto Networks acquisition, the conversion economics shift to Prisma AIRS enterprise procurement; the structural risk is that open-source community usage does not convert into commercial procurement at the rates that the standalone open-core model would have supported. The bull case is that Palo Alto Networks’ enterprise-security distribution accelerates conversion; the bear case is that open-source usage stays open-source and the commercial-revenue trajectory below open-core.
Founder-concentration and post-acquisition retention
Portkey is founder-led with Rohit Agarwal and Ayush Garg as the principal public voices. Post the May 29, 2026 Palo Alto Networks acquisition closing, the principal structural risk is founder-retention inside a much larger enterprise-security platform — the standard post-acquisition retention risk that applies to most acquired founder-led startups. The bull case is that the Prisma AIRS AI Gateway anchor positioning gives the founders structural visibility and product-leadership scope; the bear case is that integration friction or post-acquisition incentive structures drive founder-attrition over the multi-year horizon.
Recent IM Coverage
- AI Infrastructure — sector landing May 2026.
- AI Tracker — methodology and universe May 2026.
Show recent press coverage of Portkey
- May 2026 — Palo Alto Networks Completes Acquisition of Portkey to Secure AI Agents (Palo Alto Networks)
- Apr 2026 — Palo Alto Networks to Acquire Portkey to Secure the Rise of AI Agents (Palo Alto Networks)
- Mar 2026 — Portkey’s Gateway is Now Fully Open Source, Processing over 1 Trillion Tokens Every Day (Portkey Blog)
- 2024 — AI apps building platform Portkey raises $15 Mn in Series A led by Elevation (Entrackr)
- 2024 — Portkey Raises $15M Series A to Scale the Unified Control Plane for Production AI (Portkey Blog)
- Aug 2023 — Portkey.ai raises $3m to accelerate generative AI apps (Ventureburn)
Show the source register for the figures on this page
IM operates a primary-source-where-possible discipline. The figures above come from:
- Usage — tokens and organisations: Portkey’s March 2026 open-source-Gateway announcement disclosed processing 1T+ tokens / day and 120M+ AI requests / day across 24,000+ organisations including Fortune 500 customers across finance, pharma and technology. We anchor to the 1T+ tokens / day and 24,000+ organisation figures for early-2026 as the principal usage signal.
- Revenue: Portkey does not separately disclose precise ARR figures in primary filings as of the May 2026 Palo Alto Networks acquisition close. The principal commercial signal is the AI-spend-under-management figure of $180M+ in annualised AI spend per the March 2026 open-source-Gateway announcement; we decline-to-publish a precise ARR figure and reference the AI-spend-under-management signal as the canonical commercial-scale indicator. Post-acquisition revenue contribution to Palo Alto Networks Prisma AIRS will be disclosed inside Palo Alto Networks’ standard 10-K and 10-Q segment-reporting cycle.
- Headcount: Portkey does not publicly disclose precise headcount in a primary filing as of the May 2026 Palo Alto Networks acquisition close. LinkedIn-visible data places the company in the low-hundreds range as of the acquisition cycle; we decline-to-publish a precise figure and reference the company’s careers and about pages as the canonical entry point. Post-acquisition headcount integrates into Palo Alto Networks’ overall employee base.
- Funding to date / acquisition: Cumulative external venture capital approximately $18M through the 2024 $15M Series A led by Elevation Capital with Lightspeed participation, plus the August 2023 $3M seed led by Lightspeed per Ventureburn. Acquired by Palo Alto Networks May 29, 2026 per Palo Alto Networks’ completion press release; specific acquisition deal terms not separately disclosed in the press cycle.
Methodology & Disclaimer
For metric definitions, source-tier hierarchy, and decline-to-publish rules, see the tracker methodology. Confidence dots (• green / • amber / • red) follow the same convention as the AI Tracker.
Spotted a figure you believe is wrong? Send corrections to info@informationmatters.net.
Information Matters Framework scores are the considered opinion of the IM team — human and AI — applied to publicly-available evidence under a disclosed methodology. They are not statements of fact about the companies scored and they are not investment advice.
