Repository navigation
Home
Your agent pays for its tool definitions on every request. toolism measures that cost and cuts it, without changing which tool calls are valid.
Every tool you give an LLM (an OpenAI function, an Anthropic tool, an MCP server tool) is sent as JSON Schema with every request. Schema generators pad that JSON with things the model already knows. toolism removes exactly that, checks that the result accepts the same tool calls, and gives you a CI gate so the cost can't creep back up.
On 14 real tool sets (159 tools from 13 popular MCP servers) the default safe rules cut tool-definition tokens by 37.4% in total (see the benchmark).
| Page | For |
|---|---|
| Getting Started | Install it and run your first measure / slim / check |
| CLI Reference | Every command, the common options and exit codes |
| Python API | Using toolism from agent code |
| Rules | What each rule removes, and the safety levels |
| CI Integration | GitHub Action and pre-commit hook |
| How Verification Works | Why the slimmed tools accept the same calls |
| Contributing | Picking an issue, opening a PR, how review works |
| FAQ | Prompt caching, accuracy, model behaviour |
| Roadmap | What's planned and where help is wanted |
The README and the files in the repository are the source of truth; these pages are the quick guide and link to them for details.