Conversation history
AI Leash keeps your conversations on your computer so you can return to them later. Your history is still there after you close and reopen the app, and conversations from different projects stay separate.
What is saved
As you chat, AI Leash saves the conversation as it develops:
- Your messages and the agent’s replies
- Tool activity and its results, such as files the agent reads or commands it runs
- The order and time of the messages
- How long a completed agent reply took, when that information is available
Saving happens during a turn rather than only at the very end. If the app closes unexpectedly, the conversation can therefore retain the part of the exchange that was already received.
The app does not save the agent’s temporary streaming “thinking” display. It also rebuilds the instructions supplied by your project context when you use the conversation again, rather than showing those instructions as part of the visible transcript. See Context and memory for how project context is provided to the agent.
What “across restarts” means
When you reopen AI Leash, your saved conversations are available in the conversation list. Select one to restore its transcript and continue where you left off. Message timestamps, tool activity, titles, and completed-turn durations are restored with it.
History is stored on the computer running AI Leash. It is not dependent on keeping the app window open, and it is not lost simply because you switch to another project or start a new conversation.
Finding and managing conversations
The left sidebar lists your conversations, with the most recently active ones first. It normally shows conversations from all projects. Use the project filter at the top of the sidebar to narrow the list to one project.
Select a conversation to switch to it. Switching does not delete or merge anything: each conversation has its own transcript and can be continued independently.
The tabs you had open in a conversation, in the side panel and for its sub-agents, are remembered with it. They come back when you reopen the conversation, including after restarting the app. See Tabs are remembered per conversation.
To start a separate conversation, use New Thread and choose the project you want to work on. You can have multiple conversations for the same project—for example, one for investigating a bug and another for planning a feature.
Nothing is saved for a new thread until you send its first message: it isn’t listed in the sidebar, no agent session is started for it, and its text, model and worktree choices are dropped if you leave without sending.
AI Leash gives conversations a title based on the beginning of your first message. Some agents can provide a more specific title. The title helps you recognize the conversation in the sidebar; it does not change the transcript.
When you are finished with a conversation, mark it Done to move it into the collapsed Done section. You can reopen it or mark it as not done later. To remove a conversation permanently, open its context menu in the sidebar and choose Delete conversation. Deleting removes its transcript and its related sub-agent records as well.
Board and task chats
Each project’s board is backed by its own project chat, saved like any other conversation but opened from Project boards in the sidebar rather than listed under Chats. Removing a project from AI Leash deletes its board and its project chat with the rest of its conversations.
A started task’s chat is a normal conversation, listed under Chats with the task’s number (or grouped or hidden there — see the filter next to the project menu). Moving the task to Done marks its chat done too. Deleting a task leaves its chat; deleting the chat leaves the task on the board, where it can be started again.
Messages and notes from other chats are saved in a conversation’s transcript with their sender, so they come back after a restart like everything else.
Sub-agent work
When the main agent delegates a task, the sub-agent appears in the Sub Agents view for the current conversation. Each task shows its description, status, model, and how long it has been running or ran.
Select a sub-agent to open its transcript after it finishes—or while it is still running—and review the work it performed. Its prompt, messages, tool activity, and final result are kept so you can inspect the work later, even after restarting the app. A completed or failed sub-agent can be deleted from the Sub Agents view; a running one can be stopped.
Sub-agent conversations are kept with the conversation that created them. Deleting or clearing that parent conversation also removes the associated sub-agent history.
For more about delegated work and how it appears in a chat, see Agent chat.