Skip to content

tzachbon/srulik-toolkit / plugins/mcporter-plugin/skills/mcporter-setup

mcporter-setup

Unverified

Move the MCP servers configured in Claude Code, Codex, Cursor and other editors into one mcporter home config, check each one answers, then disable the native copies with backups and an undo receipt. Use when the user asks to set up mcporter, consolidate or migrate MCP servers, or stop using native MCP tools.

mcporter setup

Goal: every MCP server the user has configured is reachable through mcporter call, and the harness no longer loads a second native copy.

Run the work in a child agent when the host offers one (the mcporter-setup agent in this plugin). Otherwise follow these steps yourself. Either way, the user approves step 3 before anything is written.

Steps

  1. Inventory. Run node <plugin>/scripts/setup.js inventory. It lists the servers mcporter can import from every supported editor config, and the servers already in mcporter's config. It prints names and paths, never definitions, and lists any config file it could not read under unreadable.
  2. Classify each server and show the table to the user:
    • Move: stdio or HTTP server with a static config.
    • Move, then log in: OAuth server. After moving, run mcporter auth <name> once.
    • Stays native: servers the harness injects at runtime (the host's own tools, hosted connectors, IDE bridges). They never appear in the inventory, so nothing is done to them and the redirect hook ignores them.
  3. Ask which servers to move. Default: every Move and Move, then log in server.
  4. Dry run. node <plugin>/scripts/setup.js apply --servers a,b --dry-run.
  5. Apply. node <plugin>/scripts/setup.js apply --servers a,b. For each server it:
    • copies the definition, as mcporter read it from the first file listing the server, into the mcporter home config: the existing mcporter.json or .jsonc under $XDG_CONFIG_HOME/mcporter or ~/.mcporter, else ~/.mcporter/mcporter.json (a home config with comments is refused);
    • leaves a name alone when mcporter already holds a different definition of it (for example another project's server), listed under conflicts;
    • runs mcporter list <name>; only a server that answers counts as migrated. A project config/mcporter.json that defines the same name would win over the home config, so that server is listed under failed as shadowed; so is a server ordinary mcporter cannot find (for example with MCPORTER_CONFIG set). A stdio server defined for one project gets that project as its cwd. A server that needs a login is kept in mcporter and listed under authRequired: run mcporter auth <name>, then apply it again so its native copy is disabled;
    • disables every native copy of a migrated server that has the same definition, without deleting it, in the narrowest scope the harness offers: a copy defined for one project is switched off for that project only. A same-name server defined differently elsewhere, or bound to another project's folder, stays enabled and is listed under conflicts: ask the user which one to keep, or rename one. A copy the script cannot switch off safely is listed under manual or failed. Each harness's files, switches and limits are in its reference: Claude Code, Codex. Copies in any other editor go to manual;
    • checks every JSON file it edits parses before writing anything, backs up every file owner-only under the plugin data dir, ~/.mcporter/plugin (set MCPORTER_PLUGIN_DATA to an absolute path to move it; never beside a project file) before writing, and saves the receipt with its undo command after each backup. Undo refuses when a file changed after that apply, such as by a later apply, or after an apply that stopped part-way; undo the newer receipt first, or check the files and pass --force;
    • records conflicting names, so the session hint and the redirect hook leave those servers to the native tools.
  6. Report the receipt: migrated, authRequired, conflicts, failed (native copy left enabled), manual steps, OAuth logins still needed, and the undo command. Tell the user to restart the coding agent so the native copies unload.

Rules

  • Never delete a harness MCP entry. Disable only, and only after mcporter reached the server.
  • Never print secret values from configs. Refer to them by key name.
  • apply needs mcporter installed (npm install -g mcporter@0.17.0; Node 24 or newer), because everyday calls use the mcporter command. Inventory and dry runs fall back to npx -y mcporter@0.17.0.
  • After setup, daily use follows the mcporter skill.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 467kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 319kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 319kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 319kUpdated Claude CodeCodexVerifiedInstall