Armature helps companies monitor, analyze and optimize how AI agents experience their products. As MCPs, CLIs and agent-facing APIs become the new front door to software, the team argues that agent experience or AX is the next surface to instrument, the way Datadog and Amplitude instrument servers and humans today. The platform runs real agents end-to-end through a customer product and catches breakage before users do.
The product has three layers. Automated testing fires complex agent sessions against MCPs and CLIs across all major harnesses including Claude Code, Cursor, ChatGPT and Gemini, plus every leading model. Analytics and benchmarking trace each session to show what the model thought, what context it missed and why it failed, then benchmark AX against competitors. An auto-remediation MCP lets internal coding agents plug into Armature and autonomously fix AX issues based on run data, closing the loop from detection to fix.
Founded in 2026 by Louis Scremin and team, Armature is in YC Spring 2026 with backing also reported from Palantir, Tsuga and Joko. The team previously shipped MCPs used by millions and built observability products, which shows in the surprisingly mature pricing already live: Starter at $79 per month annual, Pro at $159 and custom Enterprise, with a 7-day free trial.