Wire Two MCP Servers Into a Project: a stdio Vault and a Remote Endpoint
Model Context Protocol (MCP) servers are how Claude Code reaches systems beyond the repo. In this task you author a project-scope .mcp.json with two real servers: the reference filesystem server speaking stdio, rooted at a vault of internal handbook files, and a remote keyless endpoint speaking HTTP. Then you use the vault server from inside a live session to retrieve three planted facts, and you submit the config, the session transcript, your answers, and your notes on config scopes and the trust boundary for instant, rubric-based feedback.
1 hr
Est. time
3
Outcomes
5
Rubric criteria
65%
Pass score
What you'll learn
Skills you'll have real reps in after shipping this.
The scenario
Every team that adopts Claude Code hits the same moment: the answer Claude needs lives in another system (a wiki, a tracker, a database), and pasting it into chat gets old fast. The Model Context Protocol (MCP) is the open standard that closes that gap, and the first server you wire yourself teaches the whole pattern: the two transports, the three config scopes, the approval step, and exactly what a server is allowed to see.
This kit stages that moment with the internal handbook of Otterworks, a fictional warehouse-robotics company. Six files sit in vault/; three of them hold ledger tokens that answer the kit's questions and three carry decoys. You wire the reference filesystem server over the vault as a stand-in for any external data source, add a remote keyless server alongside it so both transports are real, and prove the retrieval with a session transcript. A local self-check reads your actual .mcp.json and writes your proof-of-work evidence only when both transports are wired correctly, the config is keyless, and the retrieval artifacts hold up.
Your role
You are the engineer who owns how your team's project connects Claude Code to outside systems. Your deliverable is a shareable project-scope MCP config with both transports wired, a transcript proving the retrieval went through your server, and notes a teammate could learn the scope model from, packaged as a single submission.
Start the task to unlock the full brief
You'll get the step-by-step requirements, setup commands, the 5-criterion grading rubric, tips, and the ability to submit your solution for instant AI grading.
Free to start · submit when you're ready
Learning resources
Wire real MCP servers into Claude Code, hands-on
The Model Context Protocol (MCP) is how Claude Code reaches wikis, trackers, databases, and every other system your answers live in, and this task has you wire it for real. You author a project-scope .mcp.json with both transports (a stdio filesystem server rooted at a vault of company documents, plus a remote HTTP endpoint), approve the servers in a live session, and drive the retrieval of three planted facts through namespaced mcp__ tool calls you capture in a transcript. Along the way you learn the three config scopes and why the project-scope file is the one a team shares, what a server can see given its root argument, and why a committed config stays keyless. A local self-check verifies the wiring against your actual config and generates your evidence, so a green run means a submittable bundle.