Alloy Robotics, a startup using AI agents to diagnose failures across robot fleets, has raised $8 million at an $80 million valuation, little more than a year after its founding.
The funding round was led by Square Peg, with existing pre-seed investors Blackbird, Airtree and Skip Capital participating again. The round also drew investment from executives and engineers associated with OpenAI, Anthropic, Tesla (NASDAQ: TSLA), Waymo, Halter and Carbon Robotics, as well as several Alloy customers.
The funding comes as robotics companies deploy larger fleets across environments ranging from warehouses and farms to defence, maritime operations and construction. As fleet sizes grow, engineers face a corresponding increase in the volume of logs, telemetry, video and sensor data generated by machines.
Alloy is targeting that problem with software designed to identify the causes of robot failures and recurring problems.
Turning robot data into engineering insight
Robot failures can require engineers to examine multiple sources of information before determining what went wrong. Alloy brings those sources together and adds engineering context from tools such as Slack and Jira.
Its AI agents look for anomalies, regressions and recurring patterns. The platform then links findings to specific missions, timestamps and signals, giving engineers a way to trace problems back to the underlying data.
“When a robot fails, an engineer can spend days, sometimes weeks, working out why,” said Joe Harris, founder and CEO of Alloy Robotics. “The answer’s in the data, just buried.”
Harris previously served as chief commercial officer at Eucalyptus, which was acquired for $1 billion.
The company said its platform now supports close to 1,000 robots and has analysed more than 10,000 missions, with most of that activity occurring during the past two months.
Cutting analysis time
Alloy’s customers are using the platform to reduce the time spent investigating field failures and validating robot performance.
At Advanced Navigation, field-test analysis that previously took about a day can now be completed in less than 10 minutes, according to the company. The platform also helped the team complete 44 field tests in slightly more than a day after work that had previously taken weeks.
Jai Castle, product validation manager at Advanced Navigation, said the change has allowed the team to focus on additional testing rather than whether existing work can be completed on schedule.
The platform is also being used by DroneForge, a US autonomous-drone company. In one case, engineer David Crabtree suspected that a particular component was responsible for a failure. Alloy’s analysis showed that both state estimators were operating normally and pointed the investigation towards the actual fault.
Such diagnosis can be important for autonomous systems, where repeated misdiagnosis can delay fixes and allow problems to compound across a fleet.
Connecting AI coding agents to robot data
Alloy is also integrating its platform with AI-powered coding tools. Through a native Model Context Protocol (MCP) server, coding agents such as Codex and Claude Code can access the context associated with individual robot missions.
The integration allows engineers to investigate failures without manually collecting and organising data from multiple systems.
Alloy is currently used across several robotics categories, including navigation, defence, drones, agriculture, maritime systems, humanoid robots, construction and medical robotics.
Jethro Cohen, principal at Square Peg, said robotics companies that can learn quickly from operational data will have an advantage as fleets expand.
Funding to support expansion
The new capital will be used to expand Alloy’s engineering team, support its US operations and develop its AI models and agent platform.
The company’s rapid fundraising reflects growing investor interest in software infrastructure for physical AI. While much of the robotics industry focuses on building machines and improving autonomy, fleet operators also need systems that can monitor performance and identify failures once robots are deployed.
For Alloy, the opportunity lies in turning those operational records into a continuous feedback loop for engineering teams.
“Getting a robot to work is only the beginning,” Harris said. “To earn trust at scale, teams need to learn from every run.”
With its latest funding, Alloy is betting that as robot fleets become larger and more complex, the ability to understand what those machines are doing—and why they fail—will become as important as the robots themselves.
