Settings

This page covers user-facing settings and safe configuration practices in TatsuCode.

Where Settings Are Stored

TatsuCode keeps settings and credentials in your local user data area.

• Settings file: config.json
• Credentials file: credentials.enc
• Session/project data: stored locally per user profile

For security reasons, this documentation intentionally avoids exposing low-level storage internals.

Core Settings You’ll Use Most

Theme

/theme

Switch visual themes for readability and preference.

Model Selection

/models

Pick the right model for speed, quality, or context size.

Provider Connections

/connect

Configure OpenRouter, OAuth subscriptions, and local providers.

Temperature

/temp 0.7

Set response creativity:

• Lower = more deterministic
• Higher = more creative

Reasoning Controls

/reason-effort
/reason-display

Control how deeply supported models reason and whether reasoning output is shown.

Context Management

Compacting Long Sessions

/compact

Use this to keep long sessions responsive and within context limits.

Options include:

• Auto compact behavior
• Manual compact now

Fresh Session

/new

Start fresh when changing tasks or project focus.

Permissions

/permissions

Choose how TatsuCode handles sensitive actions:

• Ask each time
• Allow in trusted contexts
• Deny by default

Use this to balance speed and safety.

Project-Level Instructions (AGENTS.md)

/init

/init analyzes your repository and creates or updates AGENTS.md in the project root.

It is intended to capture practical project context (commands, architecture, and conventions), not just generate a blank template.

Reference: agents.md

Use AGENTS.md for baseline project behavior that should apply every session.
See detailed guide: AGENTS.md

Reusable Specialized Workflows (Agent Skills)

Use skills for focused workflows that are too specific for AGENTS.md.

/skills
/skills-template
/skills-refresh

Reference: agentskills.io

Good pattern:

• AGENTS.md = broad project defaults
• Skills = targeted playbooks (e.g., release audit, deep debug, migration checklist)

PATH Integration

Add launcher

/add-tatsu

Remove launcher / cleanup

/remove-tatsu

Usage Monitoring

/usage
/usage-quota

Use these to monitor token usage and provider limits.

Security Best Practices

• Keep your machine user account protected
• Do not share credentials.enc
• Rotate API keys periodically
• Disconnect providers you no longer use
• Review generated diffs and commands before approval

Next Steps

• Providers — provider connection guide
• Models — model strategy and selection
• AGENTS.md — project context file guide
• Agent Skills — full skills workflow
• Commands — complete slash command reference