Server facts
token-pilot MCP Server
Project
- Author
- Digital-Threads
- Repository
- github.com/Digital-Threads/token-pilot
- License
- Not specified
- Category
- Developer Tools
Tech
- Runtime / language
- node
- Transports
- stdio
- Remote available
- No
- Authentication
- Not specified
Activity
- GitHub stars
- 5
- Last commit
- Data as of
Tags
- Cursor
- Python
- Typescript
- ast
- claude code
- code reading
- codex
- developer tools
- mcp
- token savings
Overview
Repository exploration in token-pilot centers on code structure rather than repeated full-file reads. project_overview identifies architecture and tooling, while outline and smart_read expose symbols for closer inspection with read_symbol. Developers preparing changes can use read_for_edit to obtain raw source around selected symbols or lines, optionally including callers, tests, and changes. After a baseline has been captured with smart_read or read_for_edit, read_diff returns the edited hunks. Documents and data have a separate path: read_section selects content by heading, top-level key, or CSV row range.
For debugging and refactoring, find_usages groups references, call_tree follows call relationships, and related_files ranks connected source and tests. Module tools inspect dependencies and trace import paths; smart_diff connects Git changes to affected symbols. Test execution produces structured results through test_summary. Session analytics report tool activity and estimated token savings, while session_budget describes read-hook pressure rather than context-window occupancy. Alongside its MCP tools, the project provides hooks that redirect heavy native reads and searches, plus Claude Code-specific subagents with response budgets.
token-pilot setup notes
- token-pilot records stdio only; confirm a network transport upstream before planning remote access.
- token-pilot has no recorded license; confirm reuse terms in its repository.