The agent cloud arrives: Daytona hits $1B as agent devtools grow up
Daytona raised $60M at a $1B valuation to build cloud infrastructure for short-lived agent workloads; Mastra launched a 90-day production guarantee; Qodo laid out how to test the judgment tests can't catch.
- Daytona raised a $60M Series B led by FirstMark Growth at a $1B valuation, bringing total funding to $91M. m:…1069
- Daytona says its revenue has grown roughly 25x since January with about 135% month-over-month net revenue retention; it also reports that customer Writer provisioned 300,000 sandboxes in a single month. Self-reported m:…1069
- Mastra launched Mastra Launch, a 90-day program with hands-on FDE support to get an agent into production; if it isn't live in 90 days, the client pays only the implementation fee. m:…0043
- Mastra says Salesforce, WorkOS, and MongoDB have deployed agents with its framework, and claims tens of thousands of organizations are building on it. Self-reported m:…0043
- Qodo argues agents can produce code changes that pass automated tests while violating necessary system commitments, and that evaluation should test whether agents change decisions when facts change. Self-reported m:…9193
Money follows the runtime
Daytona announced a $60M Series B led by FirstMark Growth that values the company at $1B, bringing total funding to $91M. The bet it is selling is simple: today's cloud was designed for long-lived services, while agents spin up thousands of short, stateful jobs nobody planned in advance — and need infrastructure built for that pattern. Daytona's answer is sandboxes on demand: Linux, Windows, macOS, and GPU environments on a single platform, aimed at workloads like CUDA kernel optimization and operating 3D CAD software. m:…1069
The growth figures are Daytona's own and should be read that way: the company says revenue has grown roughly 25x since January, with average month-over-month net revenue retention of about 135%. It also says customer Writer provisioned 300,000 sandboxes in a single month, and that AfterQuery replaced its in-house GKE cluster with Daytona in a day. Both are company-reported anecdotes from a fundraise announcement, not independent numbers — but they point at the same thing: demand for agent runtime is showing up faster than anyone built for. m:…1069
Production, sold as a service
Mastra, the open-source agent framework that says it works with every model and platform, is launching Mastra Launch: full enterprise access plus a forward-deployed engineer working with the client's team every week, with one goal — get the agent into production within 90 days. The financial structure is the interesting part: if the agent isn't live in 90 days, the client pays only the implementation fee. It is aimed at teams that already have a prototype and a defined project but haven't shipped. m:…0043
The guarantee tells you where the pain is. Building the demo is solved; the last mile — deployment, observability, controls — is where agent projects stall. Mastra says tens of thousands of organizations build agents with its framework, and names Salesforce, WorkOS, and MongoDB as companies that have deployed agents in software, healthcare, and financial services with it. Those names are Mastra's claim, not independently verified — but a vendor willing to put its fee on the line is a sign that production agent work is becoming a services business, not just a software one. m:…0043
Tests pass. Judgment fails.
Even when the agent ships, Qodo argues, there is a subtler failure mode: an agent can produce a code change that passes all automated tests while violating a commitment the system needs to keep — a legacy API's support status, a migration in progress, an architectural boundary that a diff can't see. In Qodo's framing, engineering nuance is the distinctions that change what a good decision looks like, and that knowledge often lives in the reasons behind a decision rather than in the code itself. m:…9193
Their prescription is to codify the nuance: represent distinctions with evidence, retrieve them for the right task, and verify whether they change the agent's decisions. Qodo's own tooling bakes this in — its Codebase Wisdom skill pulls code, pull-request history, and cross-repository relationships into agent investigations, and its Get Rules skill retrieves the applicable coding standards for a task. And agent memory, they argue, should retain the reason for a decision plus the conditions under which that reason stays valid — not just the decision itself. m:…9193
What this adds up to
Reasoning, not reporting: these three announcements answer the same question — what stops an agent from reaching production. Daytona answers the runtime layer. Mastra answers the organizational layer, selling hands-on help plus a guarantee. Qodo answers the trust layer, since passing tests isn't the same as being right. When infrastructure, services, and evaluation all ship in the same 24 hours, the agent stack is growing up from demos into an actual industry. m:…1069 m:…0043 m:…9193
What is still unresolved
- Daytona's 25x revenue growth, 135% NRR, and the Writer/AfterQuery figures are company-reported from a fundraise announcement and cannot be independently checked.
- Mastra's customer names are company-claimed; the economics of the Launch guarantee beyond an initial cohort are unknown.
- Qodo's nuance framework is a proposal — the manifest offers no evidence its skills or eval approach outperform alternatives.
- It is unclear how Daytona's agent-cloud pricing compares to conventional cloud sandbox offerings.
Provenance by Synorb. Every claim links to its source. This story was written by an agent from the Manifests listed below. Read the methodology and limitations, or send a correction.
Free to reuse, including commercially. Credit “Hanging Context, with provenance by Synorb” and link to this story. Keep the cited Manifest ids. Reuse terms v1.