tokensave

Build a code-graph MCP server for symbol search and call-graph analysis.

Updated May 19, 2026
One-click install
npx skills add https://github.com/blouargant/yoke-registry --skill tokensave
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: tokensave
Source: https://github.com/blouargant/yoke-registry/tree/main/skills/Coding/tokensave
Command: npx skills add https://github.com/blouargant/yoke-registry --skill tokensave

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Tokensave eliminates slow, repetitive code reading by building a searchable code graph so Claude can answer structural questions about a project quickly and accurately.

Core Features & Use Cases

  • Code-graph MCP setup: Connect Tokensave to Claude via MCP so the model uses graph-based tools instead of raw file reads.
  • Project indexing and syncing: Initialize a database from the project source and keep it current with incremental or forced re-indexing.
  • Deep code analysis tools: Enable symbol search, call-graph exploration, impact/affected analysis, and code quality checks.

Use Case: You are refactoring a large codebase and need to understand how a change to a function impacts callers, tests, and architectural coupling without manually grepping through thousands of files.

Quick Start

Run tokensave claude-install and then start an MCP server pointing to your project root so Claude can use tokensave tools for symbol search and call-graph questions.

Frequently Asked Questions about tokensave

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How does a code-graph MCP server help with impact analysis in a large codebase?

A code-graph MCP server performs impact analysis by indexing your project source so Claude can use graph-based tools instead of raw file reads. This allows it to quickly identify how a function change impacts callers, tests, and architectural coupling without manual grepping.

How do I set up tokensave to enable symbol search and call-graph exploration in Claude?

To enable symbol search and call-graph exploration, run the tokensave claude-install command to configure the MCP connection. You then start the MCP server pointing to your absolute project root directory so Claude can access the indexed code graph tools.

Can I maintain per-branch code graphs for consistent results across different checkouts?

Yes, tokensave supports maintaining per-branch code graphs for consistent results across checkouts. By initializing and syncing separate project indexes for each branch, you ensure accurate structural queries and impact assessments during refactoring.

What's the best way to keep my project index current when refactoring a large codebase?

The best way to keep your project index current during refactoring is using incremental syncing or forced re-indexing. This updates the searchable code graph database with the latest source changes, ensuring Claude's structural queries and code quality checks remain accurate.

Why does the tokensave MCP server require an absolute path to the project directory?

The tokensave MCP server requires an absolute path to the project directory to correctly locate and serve the initialized code graph database. Serving with the --path flag ensures Claude reliably resolves the index for symbol search and call-graph analysis workflows.

Do I need to manually read files before starting a code exploration workflow with tokensave?

No, you do not need to manually read files. Tokensave eliminates repetitive code reading by using the tokensave_context tool as the starting point for exploration workflows, allowing Claude to answer structural questions directly from the indexed code graph.