Docs

Chat & Sessions

The Chat is where you talk to the orchestrator. Every conversation lives inside a named Session, so you can switch contexts without losing history — one session per feature branch, sprint, or topic area.

Sessions

Each project can have multiple named sessions running in parallel. Example sessions for a single project:

Feature: AuthWorking on OAuth login flow.
Bug sprintTriage and fix open bugs from the issue tracker.
GeneralOpen-ended questions and architecture discussions.

Create a new session with the + button at the top of the sessions list, or start one automatically when wrapping up the current session.

Wrap up a session

Use /wrapup in chat or click the Wrap up session button on the Dashboard. The orchestrator generates a markdown summary of everything that happened:

→Completed agent tasks and their outcomes
→Issues closed or status changes
→Files modified and git commits made
→Errors encountered and how they were resolved
→Suggested next steps for the next session

If memory is configured, the summary is automatically saved to your Obsidian vault or NotebookLM for retrieval in future sessions.

Init context

Use /init in chat or click the Init contextbutton on the Dashboard at the start of a session. Agenties reads from your vault or NotebookLM and generates a briefing file that gives the orchestrator context about the project's history and current state — what was built, what broke, what's pending.

Tip:Run /initat the start of every work session. It takes 10–20 seconds and dramatically improves the orchestrator's first response quality.

Session history

Type /history in chat to browse past sessions. Each entry shows the session name, date, and its wrapup summary (if one was generated). Click a session to view the full conversation.

Export

Export the current chat for archiving or sharing:

/export markdownSaves the conversation as a .md file with timestamps and role labels.
/export jsonSaves the full structured message array as JSON, including metadata.

Composer features

The message input at the bottom of the chat panel supports several power features:

Slash commands

Type / to see available slash commands. See the Slash Commands reference for the full list.

Model and effort selection

Click the model badge next to the send button to change the orchestrator's model or effort level for the next message. Changes persist for the session.

Voice input

If voice is configured (Settings → Voice), hold the push-to-talk key to speak your message. Whisper transcribes it and inserts the text into the composer. See the Voice Chat page for setup instructions.

Artifacts

Some agent responses include rendered artifacts — HTML pages, SVG diagrams, or formatted markdown reports. A view artifact button appears in the chat message. Clicking it opens the artifact in a slide-in drawer next to the chat. See the Artifacts page for details.