Agentic Chat (with Approvals)

Chat that can act,without surprising you

A safe, mode-based AI chat that can create notes and tasks with human-in-the-loop approvals. Ask questions in Ask mode, edit content in Edit mode, and run workflows in Agent mode.

Try it in the editorAll features

Why this exists

AI should follow instructions, not guess your workflow

Traditional “agent” chats can misfire: you ask a question and it starts creating a whole hierarchy. Agentic Chat fixes that with explicit modes and approvals.

Safe by default (Ask mode)
Ask mode is Q&A only—no task extraction, no note creation, no background workflows.
Approvals for every create
In Agent mode, the assistant proposes actions in chat. You approve each item individually before it becomes a real note.
Provenance built-in
Created notes can include a “Derived from” relationship under the title and store the relationship in the database.
No accidental side effects
Agent mode proposes actions in chat, but nothing is created until you approve.

Modes

Three modes, one clear intent

The mode switch sits right next to the input (like Cursor) so it’s always explicit what the assistant is allowed to do.

Ask

Q&A only. Great for “Where is ICP draft mentioned?” and other questions.

  • No creation workflows
  • No accidental proposals

Edit

Suggest edits to the current note, then apply them in one click.

  • Applies structured content
  • Nested list formatting supported

Agent

Runs workflows (create notes/tasks) with approvals for every item.

  • Single-note vs multi-note aware
  • Tasks only when you say “tasks/todos”

Smart Context

AI that understands your full knowledge base

Control exactly what context the AI sees—from semantic search across all notes to precise @mentions of specific notes.

Semantic Search Context

Toggle on to automatically include relevant notes as context. Uses AI embeddings to find semantically related content across your workspace.

  • Finds related notes automatically
  • Works across your entire workspace

Daily Updates

Ask “What did I work on today?” or “Summarize this week” and get instant summaries of your recent activity.

  • Today, yesterday, this week ranges
  • Auto-detected from your question

Note Mentions

Use @mentions to include specific notes as context. Perfect when you want precise control over what the AI sees.

  • Type @ to search notes
  • Combine multiple notes as context

Image Attachments

Attach screenshots, diagrams, or photos to your messages. The AI can analyze and discuss visual content.

  • Drag and drop or paste images
  • Multimodal conversations

Workflows

From transcript to next steps in minutes

Agentic Chat supports common PM flows: turning meeting context into a structured note, extracting tasks, or breaking a large note into multiple linked notes.

Create a single note (approval required)

Say “create a new note…” and you’ll get exactly one proposal card to approve.

Create tasks (only when explicitly requested)

Task proposals run only when you use words like “tasks” or “todos”. Each task requires approval.

Example prompts

“Create a new note that outlines next steps based on this meeting transcript.”
“Create tasks from this note.”
“Break this down into notes.” (creates multiple notes, each approved individually)
“Create a new note that summarizes this, then create tasks from it.”

Context-aware prompts

“What did I work on today?” (uses daily updates context)
“Summarize everything related to Project Alpha” (uses semantic search)
“Compare @Meeting-Notes with @Product-Spec” (uses note mentions)
“What's in this screenshot?” (with image attached)

Ready to ship faster with safer AI?

Use Ask for clarity, Edit for precision, and Agent for action—with approvals on every change.

Open EditorBrowse Features