Explore the .anote Directory¶
Panacea's CLI reads configuration from two places: a per-project file and a global one.
Project config¶
Panacea searches upward from your current directory for the first file it finds, in this order:
.anote.json.claw.jsonanote.config.json
json
{
"model": "claude-sonnet-4-6",
"permissionMode": "default",
"maxTurns": 20,
"compactAfterMessages": 40,
"hooks": {
"preToolUse": [],
"postToolUse": []
}
}
| Key | Purpose |
|---|---|
model |
Default model for this project |
permissionMode |
default, acceptEdits, or bypassPermissions — see Permission modes |
provider |
Explicit provider override (usually auto-detected from model) |
baseUrl |
Base URL for OpenAI-compatible endpoints, e.g. http://localhost:11434/v1 for Ollama |
maxTurns |
Turn cap per session |
compactAfterMessages |
When to compact session history |
hooks |
preToolUse / postToolUse shell hooks — see Extend Panacea |
anote init creates .anote.json for you. anote config reads and writes it:
bash
anote config # show effective config (global + local)
anote config get model
anote config set model gpt-4.1
anote config path # print the global config file path
anote config edit # open the global config in $EDITOR
Global config¶
~/.anote/config.json holds your defaults — applied whenever a project doesn't override them. Project config always wins over global config.
CLAW.md¶
Not JSON — a markdown file the agent reads for project context at the start of every session. See Extend Panacea for what goes in it.