questions people ask about each exam — difficulty, cost, format, prerequisites, renewal, and how to pass.
It's Anthropic's foundational certification for building on Claude — the first AI-company-specific architect credential, launched in 2026. It validates designing, building, and operating production systems across the Claude API, Claude Code, and the Model Context Protocol (MCP), including agentic workflows, tool design, prompt engineering, and context management. The exam is 60 multiple-choice questions in 120 minutes with a 720/1000 passing score. See the CCA-F hub for the full domain breakdown.
It's open to everyone — it's a public, foundational certification with a standard exam fee, not a partner-only or invite-only program. Anyone building with Claude (developers, AI engineers, solutions architects) can register and take it; you don't need to be an Anthropic partner or customer of a particular tier.
It's positioned as foundational, but "foundational" here still means hands-on familiarity with the Claude ecosystem — you're expected to understand agentic orchestration, Claude Code configuration, MCP tool integration, and prompt/structured-output patterns, not just Claude trivia. People who've only used the API casually tend to find the Claude Code and MCP domains challenging. Drilling practice questions is the best way to gauge readiness.
60 multiple-choice questions in 120 minutes, with a passing score of 720 out of 1000. That's a comfortable two minutes per question, and there's no penalty for wrong answers, so answer every one.
It's not free — the exam costs $99 USD. Anthropic provides free study material (documentation and guidance), but the exam itself is paid. See our Anthropic exam info page for registration details.
You register through Anthropic's certification portal and take the exam online. Part of the enrollment involves an access step (agreeing to the exam's terms and confirming your details) before you can sit it. Our Anthropic exam info page walks through the registration process and requirements.
That's part of Anthropic's enrollment flow — before taking the exam you complete an access/attestation step, confirming your identity and agreeing to the certification's terms and conduct policy. It's a standard gating step in the registration process, not a separate approval or partner requirement, and it's how Anthropic grants you access to sit the exam.
Get genuinely hands-on with the Claude ecosystem — build agentic workflows with the API, use Claude Code (CLAUDE.md, permissions, hooks, slash commands), and wire up MCP servers — then reinforce with the exam's core themes (prompt engineering, structured output, context management, orchestration). Our CCA-F question bank covers every domain with explanations, the Playbook distills the architecture patterns, and the hands-on lab reinforces real Claude workflows.
Yes, if you build AI applications with Claude or want to signal that skill — it's the first architect certification tied to a specific AI provider, it's timely as agentic and MCP-based systems take off, and it validates practical, in-demand skills. As a foundational cert it's a strong differentiator now precisely because it's new. See how it fits among AI certs on our State of Certifications report.
It's aimed at AI engineers, solutions architects, and developers who build on Claude — people designing agentic workflows, integrating external tools through MCP servers, and shipping production Claude-powered applications. It suits anyone moving from casual API use to building reliable, well-architected Claude systems.
The exam itself is 120 minutes, and total prep time depends on your Claude experience — developers already building with the API, Claude Code, and MCP can often be ready in a few weeks, while those newer to the ecosystem should budget more time to get hands-on first. The fastest path is building real Claude projects alongside working through practice questions.