Your personal system-design coach inside your AI agent — learn to design distributed systems and
prep for the FAANG system design interview the way the job actually demands it: by reviewing
flawed designs, defending tradeoffs under pushback, estimating capacity, and running mock
interviews — not by memorising reference architectures or watching the AI design the system for you.
Short targeted drills, an honest picture of your level, and mistake memory that resurfaces what you
got wrong. Your level and progress live on the RunDrill MCP server (mcp.rundrill.com), synced across
machines — not in a local file.
Why this course is different. System design has no compiler and no single right answer — a design doesn't "run", so the skill that matters is reasoning: spotting the bottleneck, naming the tradeoff, defending a choice or adapting when challenged. When an AI can draw any architecture on demand, the real risk is the illusion of competence — accepting a design that reads fine and is quietly wrong: a hidden single point of failure, a hot shard, a wrong consistency choice, a cache that lies, an unbounded queue. The signature drill hands you a plausible design with a planted defect and asks you to find it, like a senior-engineer review. Plus commit-then-defend tradeoff drills, back-of-envelope estimation, predict-the-failure-mode, and full mock interviews graded on the four interviewer dimensions with a level read (E4 / E5 / E6+).
The course climbs five levels — Foundations (scaling, latency, CAP, networking, load balancing) → Components (SQL/NoSQL, indexing, replication, sharding, caching, queues, Kafka, CDNs, APIs, WebSockets, object storage, search) → Patterns (consistent hashing, consistency models, idempotency, rate limiting, fan-out, distributed transactions, observability, security) → Systems (design TinyURL, Twitter, WhatsApp, YouTube, Uber, Ticketmaster, a payment system, and more, end-to-end) → Interview (the framework, capacity estimation, deep dives, tradeoff discussion, common mistakes, leveling). It is grounded in canonical sources (the system-design primer, DDIA-level concepts, the Hello Interview / ByteByteGo interview frameworks).
Pick your goal. Two paths, your choice: Learn system design (understand + design real systems — Foundations through full system designs), or Prep for the interview (everything plus the interview-performance layer: the 5-step framework, estimation technique, tradeoff discussion, and what E4/E5/E6+ answers look like).
Learn in your language. Interviews are usually in English, but you don't have to study only in English: set your native language and the coach explains in it while giving every term as native (English original) — so you reason naturally and still recognise the terms in the interview.
The coaching skill (skills/system-design-coach/SKILL.md) and .mcp.json are shared; each host reads
its own manifest and ignores the rest.
| Host | Reads |
|---|---|
| Claude Code / Claude Desktop | .claude-plugin/plugin.json + .mcp.json |
| OpenAI Codex | .codex-plugin/plugin.json + .mcp.json |
| Google Antigravity | plugin.json + mcp_config.json (+ rules/) |
The MCP endpoint is https://mcp.rundrill.com/skills/system-design — the skills-course host, passing
subject: "system-design". The server routes on the /skills segment and ignores the course name;
the name makes this register as its own MCP server in your agent. On first use the host opens a browser
tab for the OAuth handshake, then closes it — no API key to paste.
- Claude Code / Desktop — via the RunDrill marketplace:
Then run
/plugin marketplace add rundrill/rundrill /plugin install rundrill-system-design@rundrill/system-design-coach. - OpenAI Codex —
codex plugin marketplace add rundrill/rundrill, then installrundrill-system-design. - Google Antigravity — drop this folder into
~/.gemini/config/plugins/rundrill-system-design/(global) or<workspace>/.agents/plugins/rundrill-system-design/(workspace-scoped).
© RunDrill. Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) — full text in LICENSE. You may view, run, and share this plugin unchanged, non-commercially, with attribution; you may not use it commercially or publish modified/derivative versions. For other licensing, contact hello@rundrill.com.