Conversation History
Every message you send and every response the agent produces is saved automatically. Settings → History lets you browse, search, and review past conversations without losing context.
Opening History
Section titled “Opening History”- Click the Settings icon in the sidebar (or open
/settings/historydirectly). - Select History from the settings sidebar.
The history view shows conversations across all your projects and threads.
Browsing Conversations
Section titled “Browsing Conversations”The history view shows conversations grouped by thread, with the most recent at the top. Each entry shows:
- Thread name — which thread the conversation belongs to
- Project name — which project the thread is part of
- Timestamp — when the conversation took place
- Preview — the first few lines of the initial message
Click any conversation to expand it and view the full message history, including tool calls and agent responses.
Filtering
Section titled “Filtering”Use the filter controls to narrow down conversations:
- By project — show only conversations from a specific project
- By thread — show only conversations from a specific thread
- By date — filter by time range
This is useful when you want to find a previous approach you tried, or review how the agent solved a similar problem in the past.
Token Usage
Section titled “Token Usage”Each conversation entry shows token usage information, including the total tokens consumed and the model used. This helps you track API costs and understand which conversations were the most expensive.
Relationship to New Chat
Section titled “Relationship to New Chat”Starting a New Chat (Cmd+K) in a thread creates a new conversation ID but does not delete the previous one. All conversations are preserved in history. Settings → History is where you can access those earlier conversations.
Data Retention
Section titled “Data Retention”Conversation history is stored in the local Tarsk database (tarsk.db). It is never sent to external servers. Conversations persist until you delete a thread, which removes all associated conversation data.
- Use History to find solutions the agent provided in earlier sessions that you can reference again.
- When working on a long-running feature, start a New Chat for each logical step to keep conversations focused and the history easier to navigate.
- Review token usage in history to understand which models and tasks consume the most tokens.