Armature: Agent-experience analytics and discoverability for products

Armature is building observability and discoverability for products used by coding agents. It fits teams that need to know whether Claude Code, Codex or Cursor can actually find and use their product.

What it does

Armature’s homepage ↗ splits the product into two sides. Discoverability is a managed service that measures how coding agents choose tools and works with the company on docs, SDK and content changes. Usability is self-serve MCP analytics and evals: teams install the SDK, replay sessions, group use cases and catch regressions.

The product is specific about the unit that counts. Armature says a product is “picked” when an agent installs it and wires it into a repository, not when the agent merely mentions it. That is a more useful boundary than treating a page impression or a name in an answer as adoption.

Pricing and fit

Armature’s pricing page ↗ lists the managed discoverability service from $5,000/month. The self-serve MCP analytics and evals product starts free with 1,000 monthly credits, then charges $50 per 1,000 extra credits; custom plans add support, SSO/SAML, audit logs and retention controls.

The buyer is a product company whose users increasingly arrive through agents, MCP or CLI workflows. The value is not only analytics. It is a loop from agent selection to failure trace to documentation or product change to regression evaluation.

The founders’ public context is aligned with the problem. The YC profile describes Theodore Otzenberger as a former Palantir engineer and Louis Scremin as someone who previously led AI automation and MCP work at Joko. Those are professional source facts, not proof that Armature’s measured pick rate generalizes across categories.

What could make it the wrong choice

The managed service and self-serve analytics are different products. A company that only needs to inspect its MCP sessions may not need the $5,000/month discoverability service. A company without a usable public agent surface may need to improve the product before measuring which agent chooses it.

The homepage shows example percentages, session scores and a pick-rate change. Those are product examples and company-published claims, not an independent study. I would ask about repository panel composition, model and harness versions, category controls and how simulated runs relate to actual customer traffic.

My editorial take

I would shortlist Armature for a company already exposing an MCP, CLI or agent-facing workflow and willing to change docs or onboarding based on observed failures. It is not a substitute for making the underlying product useful. The strongest product decision is the separation between being mentioned and being installed and used.

Quick facts

Sources checked

Cohort context

Armature is listed in Spring 2026. In our 2026-09-18 directory snapshot, 112 of 193 listed companies in that cohort have YC’s primary industry label B2B (58.0%). This is a current-directory comparison, not an original intake count or a performance ranking. Nine-cohort dataset.

Public website snapshot

Observed 2026-09-19 16:16 UTC in raw homepage HTML. This records visible metadata and advertised links, not agent execution or product quality.

Public observations · Collection method. Missing links here do not establish that a capability or file is absent elsewhere.