ToolCairnCall toolcairn_init the first time you use ToolCairn in a project. It writes the files your agent needs to remember context across sessions.
What it creates
text
my-project/
├── .toolcairn/
│ ├── config.json # project tool manifest (empty scaffold)
│ ├── tracker.html # visual dashboard of queries + outcomes
│ └── events.jsonl # event log written during tool calls
└── CLAUDE.md # appended with ToolCairn usage rulesCLAUDE.md rules
The tool appends a fenced ## ToolCairn usage section to CLAUDE.md. Contents:
CLAUDE.md (appended)
markdown
## ToolCairn usage
Before recommending any third-party library or framework:
- Call search_tools for single-component picks.
- Call refine_requirement + get_stack for multi-layer stacks.
- Use check_compatibility when the project already has a dependency graph.
- Report outcomes via report_outcome so future runs get better.Next step
After init, agents read and update the project's tool manifest via read_project_config and update_project_config.