Replit

Replit spent almost a decade as a browser IDE that programmers liked but investors mostly ignored. Then it bet everything on a single idea: that an AI agent could write the software, and the person with the idea would never have to touch code. That bet landed. Revenue that the company puts at a few million dollars in annual recurring revenue climbed past the $150M mark and kept going, and the valuation tripled from $3B to $9B in about six months.

Founded in 2016 by Amjad Masad, Haya Odeh, and Faris Masad, and now run out of Foster City, California, Replit describes its goal in almost missionary terms: empower the next billion software creators. Whether the pricing and the agent hold up under that kind of load is the open question, but the momentum is real.

Replit turns a plain-English description into a working, deployed application. You tell Replit Agent what you want, it plans the build, writes the code, tests its own work in a browser, fixes what breaks, and ships it, all inside a browser tab with no local setup. The pitch is that it is like having a team of engineers on demand, reachable through chat.

Underneath the chat is a full-stack platform: a collaborative cloud IDE, a built-in database, authentication, hosting, and monitoring, plus connectors to services like Stripe, OpenAI, and Google Workspace. The point is that an idea can go from prompt to live app without stitching a dozen tools together.

Building software has always required either knowing how to code or waiting in a queue for someone who does. Replit attacks both problems at once. For non-programmers, the agent removes the coding requirement entirely. For teams, it removes the ticket-and-wait cycle where sales, ops, or finance need a tool and engineering is booked out for months.

It also handles the unglamorous parts that usually stall a project: environment setup, hosting, auth, and the security review before anything goes live. The agent tests and patches its own output in a reflection loop, which Replit claims is faster and cheaper than running a general computer-use model to do the same checking.

Replit sells to two very different crowds. On one end are individual builders, students, hobbyists, and non-technical people who want to turn an idea into something real without learning to program. On the other are enterprises, where the buyers are product managers, designers, and operations leaders across sales, marketing, finance, HR, and legal who need internal tools without filing an engineering ticket.

The enterprise push is the newer, hungrier motion, complete with a dedicated Forward Deployed Engineering team that embeds with the largest accounts. Replit says users from 85% of the Fortune 500 now build on the platform.

The headline product is Replit Agent, now on its fourth generation, which plans, builds, tests, and debugs applications on its own by coordinating parallel tasks across isolated micro-VMs. It can even build other agents and automations that you trigger from Slack or email.

Everything else exists to make the agent's output real: a browser IDE for collaboration, and a bundled platform of database, auth, hosting, and monitoring so an app scales from day one. The Enterprise tier wraps all of it in SSO, SCIM, role-based access, audit logs, and a security center for teams that need governance.

Replit runs on a freemium base with usage-metered credits stacked on top. The free Starter tier gets you daily agent credits and one published project. Paid plans, Core at $20/month and Pro at $100/month, hand you a monthly pool of credits ($20 and $100 respectively) plus more collaborators, more parallel agents, and stronger models. Enterprise is a custom contract for security and controls.

The catch is baked into the model: the agent consumes credits every time it iterates, and heavy weeks can produce real bill shock. Replit calls the pay-as-you-go layer Effort-Based Pricing, which is honest about the fact that a chatty agent costs more than a quiet one.

Replit is racing the same market as OpenAI and a wave of AI coding tools like Cursor, Windsurf, and Lovable, all trying to own how software gets built now that an agent can do the typing. Its own job posts name Cursor, Windsurf, Lovable, and Claude Code as the tools its field engineers expect candidates to know, which is a fair read of the competitive set.

Replit's edge is that it is not just a code assistant bolted onto an editor. It is the whole stack, IDE, agent, database, auth, hosting, and deployment, in one place, so a non-coder can go from prompt to production without leaving the browser. That end-to-end ownership is the moat, if it holds.

The traction story is the whole story here. Replit says it grew from a few million dollars in annual recurring revenue to a $400M-plus run rate in roughly a year, with a global community of 60-plus million users. Sacra pegs annualized revenue at around $525M in April 2026, up from about $300M at the end of 2025.

On the enterprise side, one Replit account puts the jump at $25M ARR to over $150M in ten months, and the company says users from 85% of the Fortune 500 now build on the platform with more than 150,000 paying customers. These are company-stated figures, but the funding market clearly bought them.

The praise is consistent: near-zero setup, real apps from plain English, and prototyping-to-deploy that feels genuinely fast. For beginners especially, the first build is the moment that hooks people.

The complaints are just as consistent, and they cluster on money. Reviewers describe credits that burn fast, surprise bills on heavy weeks (one Reddit user reported spending $350 in a day as the agent kept iterating), and support that lags on the sub-Enterprise plans. Some also find the newest agent slower or more error-prone than older versions while still eating credits. Across a small set of G2 and Capterra reviews it lands around 3.7 out of 5: capable and beloved by many, held back by unpredictable pricing and thin support.

Replit raised $400M in a Series D in March 2026 at a $9B valuation, roughly triple the $3B it hit only six months earlier when it closed $250M in September 2025. Tracxn counts about $922M raised across nine rounds.

The cap table reads like a who's who of the AI-era investor class: Georgian led the Series D, with a16z, Coatue, Y Combinator, Prysm Capital, Craft Ventures, Google's AI Futures Fund, Databricks Ventures, and others in the mix, plus angels as unlikely as Shaquille O'Neal and Jared Leto. Management has openly targeted $1B in run-rate revenue by the end of 2026, which is the bar a $9B price now has to clear.

Replit has the rarest thing in AI right now: real, fast revenue on a product people actually use. The path to its stated $1B run-rate goal runs through enterprise, where the Forward Deployed Engineering motion and the security stack are built to win big accounts.

The risk is the thing customers keep flagging. Credit-based pricing that surprises people and support that thins out below Enterprise are exactly the frictions that stall expansion once the novelty wears off. If Replit can make the bill predictable without slowing the agent down, the $9B price looks defensible. If it cannot, the same growth that got it here gets harder to hold.

Replit is unusually blunt about how it works: "in person, on purpose." It expects people at its Foster City HQ, with several roles requiring specific in-office days, on the theory that proximity speeds decisions and sharpens the work. The stated values are builder-first and terse: ideas first, proof matters, momentum wins.

Engineering runs at high talent density on a modern cloud stack, and the company is candid enough to publish a "reasons not to work at Replit" post alongside its operating principles. It hires across engineering, sales, product, design, and go-to-market, with a heavy pull toward Foster City plus roles in New York, London, and other international markets as it expands.

Disclosed pay skews high and wide, all in US dollars per year. Engineering roles span roughly $115K to $400K across many disclosed openings, product lands in the $176K to $300K range, and go-to-market and other functions fill in beneath. The spread reflects everything from mid-level to staff and leadership hires.

Beyond base, roles advertise equity and a 401(k) with a 4% match, and enterprise sales positions are structured with commission tied to significant revenue targets. Equity and, for sellers, variable pay are a real part of the package, not an afterthought.

The registered entity is Replit, Inc., a California company, and its commercial terms make the ownership line clear: customers keep their input and own the agent's output, and Replit does not sell customer data. Enterprise deployments add SSO, SCIM, role-based access, audit logging, single-tenant environments, VPC peering, and static outbound IPs.

The security architecture is a genuine selling point, not boilerplate: isolated hardened Linux sandboxes migrating to micro-VMs, pre-deploy SAST and SCA scanning powered by Semgrep and HoundDog plus LLM reasoning, and an auto-protect system that watches published apps for new CVEs. Replit publishes a full subprocessor list and operates under GDPR, UK GDPR, and CCPA, with SOC 2 referenced across its enterprise materials.

The recent coverage tracks one arc: a company that ground for nine years, found product-market fit in AI-native building, and then raised at a pace that turned heads. The March 2026 round and the Agent 4 launch dominate the headlines, framed against the same $3B-to-$9B jump that defines the year.