Booster_Storage takes WhatChange and WhyChange as separate fields, alongside alternativesConsidered, impactLevel, testingPerformed, affectedFiles and a ticketReference — the shape of a decision record rather than a note. The thought loop is just as deliberate: isRevision with revisesThought lets a later step correct an earlier one in place, and branchFromThought with branchId keeps an alternative line separate instead of overwriting the first. The parameter names carry the whole contract here, so reading them is how you learn what each tool expects of you.
A reasoning-workflow server with four tools — a thought loop with revision and branching, a project analyser, a step planner, and a store for conclusions and the context around them.
- Booster runs a numbered thought loop: thought and thoughtNumber against totalThoughts, with nextThoughtNeeded and needsMoreThoughts governing continuation, isRevision and revisesThought for reconsidering an earlier step, and branchFromThought with branchId for an alternative line. It can reach the analyser and the step planner through Call_Booster_Analyser and Call_Booster_Steps.
- Booster_Analyser analyses a project at a given path against the user's query and the thoughts so far.
- Booster_Steps records a titled task and its steps against a project path.
- Booster_Storage keeps a conclusion with its context: what changed and why as separate fields, the technical and business background, affected files, code snippets, alternatives considered, impact level, testing performed, a ticket reference, categories, subcategories, tags and related conclusions.
A project path for the analyser, the step planner and the storage tool to work against.
One command plus a key — npm install -g mcp-booster, then supply credentials