Two days ago, I released TokenCap v1.4.0 with support for Local MCP (Model Context Protocol).
The response was great, but while using it myself, I realized something.
The difficult part wasn't MCP.
It was everything around it.
Every AI coding tool has its own configuration format and setup process. Whether you're using Claude Code, Cursor, VS Code, Windsurf, or Codex, the first few minutes usually look the same:
- Read the documentation.
- Find the configuration file.
- Copy some JSON.
- Restart the editor.
- Hope everything works.
It isn't complicated, but it's unnecessary friction.
The Idea Behind v1.5
Instead of asking developers to configure MCP manually, why not let TokenCap do it?
That's exactly what TokenCap v1.5 introduces.
tokencap mcp --init
One command.
TokenCap detects the AI coding tool you're using and writes the appropriate Local MCP configuration automatically.
Whether you're working in:
- Claude Code
- Cursor
- VS Code
- Windsurf
- Codex
…the setup process becomes the same.
Run one command and start coding.
Developer Experience First
One principle guided this release:
Developers shouldn't have to visit multiple documentation pages just to connect their AI to their repository.
The best setup process is the one you barely notice.
More Than Just MCP
While automatic MCP configuration is the headline feature, v1.5 also includes several improvements behind the scenes:
- Better repository indexing for faster intelligence generation.
- A cleaner Repository Intelligence Compiler with less duplicated data.
- Improved Local-Only security auditing and secret redaction.
- Faster builds and a more maintainable architecture.
- General improvements across the CLI.
These aren't flashy features, but they make the everyday experience noticeably better.
Still Local. Still Private.
One thing hasn't changed since the first release.
TokenCap remains local-first.
- No cloud.
- No API keys.
- No telemetry.
- Your repository never leaves your machine.
Your repository intelligence stays exactly where it belongs—on your computer.
Looking Ahead
Every release of TokenCap has been moving toward the same goal:
Making AI coding assistants understand repositories without developers constantly thinking about context.
v1.5 is another step in that direction.
Not by adding more features.
But by removing one more thing developers shouldn't have to do.
Links
Website: https://tokencap.vansharora.app
npm: https://www.npmjs.com/package/tokencap
I'd love to hear your thoughts or feedback if you give TokenCap a try.
Top comments (0)