One unified API that lets B2B SaaS products and AI agents actually do things inside the enterprise software stack.
The green blocks slide past each other and lock - which is the whole idea. Two systems that were never meant to touch, made to fit in one clean move.
You can build the smartest AI agent in the world. It can reason, plan, write poetry about your quarterly numbers. And then a user asks it to update a record in Workday, and it just... can't. That gap - between an agent that can think and one that can act - is the entire business StackOne is in.
Here is a slightly annoying truth about artificial intelligence: most of the value is unlocked not by intelligence but by integration. An agent is only useful if it can reach into the software a company already runs - the HR system, the CRM, the ticketing queue, the identity provider - and do something. Reading data, writing data, provisioning an account, closing a ticket. This is unglamorous work. It is also, apparently, worth a $20 million check from Google Ventures.
StackOne, founded in London in 2023 by Romain Sestier and Guillaume Lebedel, sells exactly this: a single unified API that connects to 200+ enterprise systems, plus a growing library of thousands of pre-built actions. Instead of a SaaS company (or an AI-agent builder) writing and maintaining a separate integration for every customer's Workday and every customer's HubSpot and every customer's Okta - a task that historically consumes engineering teams for months - they connect once to StackOne and get the whole catalog. Drata, the compliance-automation company, deployed 100 integrations on the platform. That sentence is more or less the entire product pitch.
The founders did not arrive at this idea theoretically. Sestier and Lebedel worked together for more than a decade - at Google, Oracle, and the ad-tech company Yieldify - watching integrations slow down every product they touched. When the AI-agent wave arrived, they noticed the old problem had gotten worse, not better. An agent that hallucinates a text response is embarrassing. An agent that confidently calls the wrong API in production is a liability. So StackOne built not just connectors but a proprietary tool-calling LLM: a model whose job is to help an agent find the right action and execute it correctly the first time. The company reports 91.6% first-try accuracy on its own S1 Search Bench, which is the kind of number that only matters if you have spent time watching agents get it wrong.
StackOne doesn't just reinvent integration platforms - we deliver depth, accuracy and security by default.
The word doing a lot of quiet work in that quote is "security." It is where most AI-agent demos fall apart the moment a real enterprise buyer starts asking questions. StackOne's answer is architectural rather than aspirational: by default it stores none of the customer's data, passing everything through in real time. It holds SOC2 Type II, HIPAA, and GDPR compliance. And - in a move that is either principled, strategic, or both - it open-sources its Model Context Protocol servers and its prompt-injection detection tools rather than keeping them as proprietary moats. The logic is that trust is the actual product, and trust is easier to earn when your safety tooling is out in the open.
StackOne's customers are, broadly, two kinds of people: SaaS companies who need their product to integrate with everything, and teams building AI agents who need those agents to act on something.
The commercial logic is the classic infrastructure trade. Building and maintaining integrations in-house is expensive, never-finished work - every enterprise system has its own quirks, its own auth flow, its own habit of changing an endpoint the week before your renewal. StackOne absorbs that maintenance burden and sells it back as a subscription, priced around connectors and usage. A customer connects once and stops thinking about the long tail of enterprise APIs. Drata collapsing months of integration work into 100 live connections is the version of this story that closes deals.
What makes the timing interesting is that the AI-agent wave changed who the buyer is. Integration used to be a feature request from a sales engineer. Now it is a precondition for an entire category of product. An agent that can't act is a demo; an agent that can act across HubSpot, Jira, Okta, ServiceNow, BambooHR and Workday is a product. StackOne is betting that as more companies ship agents, the value migrates away from the model and toward the connective tissue - the part that makes the model useful in a system that has compliance requirements, audit trails, and a security team that will say no.
That bet is not unopposed. The unified-API space already has established players - Merge, Finch, Apideck - and the broader integration market includes heavyweights like Workato, Tray and MuleSoft. StackOne's differentiation is aimed squarely at the agent use case: not just moving data between systems on a schedule, but giving an autonomous agent the ability to discover the right action and execute it correctly and securely in real time. Whether "built for agents" is a durable moat or a temporary head start is the open question, and it is roughly what GV is paying to find out.
StackOne is infrastructure, not an agent. It is the layer that sits between whatever you are building and the enterprise systems it needs to touch.
One API to reach 200+ systems across HR, CRM, ATS, ticketing, messaging and IAM - with thousands of ready-made actions. Connect once, cover everything.
A proprietary model that helps agents discover and execute the correct action, reporting 91.6% first-try accuracy on the S1 Search Bench.
Reach any API StackOne doesn't cover yet - build custom integrations and authentication UIs without leaving the platform.
Real-time webhook forwarding, including synthetic webhooks for systems that never natively supported them.
Spent 10+ years building SaaS products alongside Lebedel at Google, Oracle and Yieldify before starting StackOne in early 2023. Runs the company from London.
The technical half of the decade-long partnership. His line - "integrations are table stakes for B2B SaaS success, especially for AI agents" - is essentially the company's founding assumption.
The tell in StackOne's cap table is who wrote the smaller checks: angels from OpenAI, DeepMind, Microsoft and Mulesoft. The people building the frontier agents are betting on the company that connects them to the real world.
| Round | Amount | Date | Lead & notable investors |
|---|---|---|---|
| Seed | $3.6M | 2024 | Episode 1, Playfair |
| Series A | $20M | May 2025 | GV (Google Ventures) · Workday Ventures · XTX Ventures · Episode 1 · Playfair · angels from OpenAI, DeepMind, Microsoft, Mulesoft |
| Total | $24M | — | Across seed and Series A |
Illustrative depth across StackOne's published connector categories. Not exact market share.
Sestier and Lebedel launch StackOne in Q1 2023 after 10+ years building SaaS together.
Episode 1 leads a seed round to build the unified API and connector library.
Google Ventures leads a $20M Series A in May 2025, bringing total funding to $24M.
The platform crosses 1B API calls and launches a proprietary tool-calling LLM at 91.6% first-try accuracy.
"Integrations are table stakes for B2B SaaS success, especially for AI agents."
Guillaume Lebedel · Co-founder & CTO, StackOne"They're creating infrastructure for modern software and the AI agent ecosystem."
Luna Schmid · Partner, GV (Google Ventures)"StackOne integrates deeply with complex enterprise systems, unlocking more powerful use cases."
Barbry McGann · Workday Ventures"We deliver depth, accuracy and security by default."
Romain Sestier · Co-founder & CEO, StackOneThe founders worked together for over a decade - at Google, Oracle and Yieldify - before StackOne.
Angels from OpenAI, DeepMind, Microsoft and Mulesoft all backed the Series A.
StackOne open-sources its MCP servers and prompt-injection detection tools.
By default it stores none of its customers' data - everything passes through in real time.
It hit 1 billion API calls inside what it calls a $17.5 billion integration market.
It provides a single unified API and integration platform that lets B2B SaaS companies and AI agents connect to 200+ enterprise systems - across HR, CRM, ticketing, messaging and identity - and take real-time, secure actions.
It was founded in 2023 in London by Romain Sestier (CEO) and Guillaume Lebedel (CTO), who had worked together for over a decade at companies including Google, Oracle and Yieldify.
About $24M total, including a $20M Series A led by GV (Google Ventures) in May 2025 and a $3.6M seed round led by Episode 1.
StackOne is infrastructure, not an agent. It focuses on the integration and tool-calling layer - letting any AI agent reliably and securely find and execute actions inside enterprise software rather than just generating text.
Yes. It is SOC2 Type II, HIPAA and GDPR compliant, stores no customer data by default, and passes data through in real time. It also open-sources prompt-injection detection tools.