Agent Mode
Agent Mode
An in-chat tool that enables autonomous multi-step actions across your connected apps. When Agent Mode is active (via the Agent pill in the prompt bar), ZeroTwo can execute a sequence of actions — like reading a file, creating a document, sending an email, or updating a database entry — without requiring you to manually direct each step.Agent Mode uses your connected app permissions (Connectors) to perform actions. Each action is logged in the Agent Activity sidebar so you can review what was done.Related: Agent Mode documentation
Canvas Mode
Canvas Mode
A document editor embedded directly in the ZeroTwo chat interface. When Canvas Mode is active, AI responses can appear in a rich editor panel alongside the chat, rather than as plain text in the chat thread.Canvas Mode supports 7 editor types:
- Rich text (prose, formatted documents)
- Code (syntax-highlighted, language-specific)
- Mermaid (diagrams: flowcharts, sequence diagrams, ER diagrams)
- Spreadsheet
- HTML preview
- Markdown
- Math (LaTeX)
Connector
Connector
An OAuth-based integration that links ZeroTwo to an external app or data source. Connectors allow ZeroTwo to read from, write to, and take actions within connected apps.ZeroTwo supports 45+ connectors, including:
- Google Drive, Gmail, Google Calendar, Google Contacts
- Notion, GitHub, Linear, Box, Dropbox, Canva
- Supabase, and many more
Custom Agent
Custom Agent
A saved, reusable AI agent configuration built at
/agents. Custom agents let you define a specific AI persona, set of instructions, tool access, and connected app permissions that persist across sessions.Custom agents are useful for recurring workflows — e.g., a “Research Agent” that always searches the web and produces structured reports, or a “Code Review Agent” with access to GitHub.Related: Custom Agents documentationCustom Instructions
Custom Instructions
A persistent system prompt applied to all new chats automatically. Custom instructions let you define how ZeroTwo should behave, what context it should always have, or how it should format responses — without needing to repeat yourself in every chat.Set custom instructions in Settings → Personalization → Custom Instructions.Custom instructions differ from project instructions, which apply only within a specific project.Related: Custom Instructions documentation
Deep Research
Deep Research
A multi-step research engine that produces comprehensive, cited research reports on a topic. Unlike a single web search, Deep Research:
- Plans a research strategy (identifying sub-questions and angles)
- Performs multiple web crawls using Firecrawl
- Synthesizes findings across sources
- Produces a structured Canvas document with citations
Focus Mode (Search)
Focus Mode (Search)
A setting that narrows web search to specific domains or content types for more targeted results. When the Search pill is active, you can select a Focus Mode to filter where ZeroTwo looks.Available Focus Modes (8 total):
Related: Web Search documentation
| Mode | What it searches |
|---|---|
| Auto | Best sources for the query (default) |
| Company | Business and company information |
| Research | Academic papers, journals, scholarly sources |
| News | Recent news articles and publications |
| PDF documents (official reports, academic papers) | |
| GitHub | Code repositories and technical documentation |
| Personal | Your connected app data (via Connectors) |
| Professional profiles and company pages |
Memory
Memory
Auto-learned facts and preferences that ZeroTwo stores about you, derived from your conversations. Memories allow ZeroTwo to personalize responses over time — for example, remembering your preferred coding language, your name, or your area of expertise.Memories are:
- Stored per-account (not per-chat or per-project)
- Viewable and deletable from Settings → Memory
- Limited to 5 memories on the Free plan; higher on paid plans
Model Picker
Model Picker
The interface for selecting an AI model in the chat prompt bar. The Model Picker is accessible via a dropdown in the bottom-left of the chat input area. It shows all available models organized by provider, with indicators for premium vs. standard status.You can switch models mid-conversation — each message can use a different model if desired.Related: Model Picker documentation
Premium Model
Premium Model
Project
Project
An organizational container for related chats, files, and instructions. Projects let you keep different workflows separate — for example, one project for a client engagement, one for personal research, and one for coding work.Each project can have:
- Its own set of uploaded files (accessible in all chats within the project)
- Project-specific custom instructions
- A collection of related chat threads
Skill
Skill
A lightweight, reusable AI tool that can be invoked in any chat using
@skill-name. Skills are purpose-built AI tools for specific tasks — similar to functions or plugins — that you can call on demand.Skills can be built and customized at /agents and may include custom logic, specific model configurations, or tool permissions.Related: Skills documentationStandard Model
Standard Model
A non-rate-limited AI model that has no monthly cap. Standard models are available on all plans (including Free) and do not consume your premium message quota.Standard models include:
- GPT-5-mini (OpenAI)
- GPT-4o-mini (OpenAI)
- Gemini Flash Lite (Google)
- Mistral Small (Mistral)
- Grok 4 Fast (xAI)
Studio
Studio
Dedicated creative generation pages at
/studio/* for generating media beyond text:/studio/images— text-to-image generation/studio/video— text-to-video generation/studio/audio— text-to-audio/speech generation
Task
Task
A scheduled AI job that runs automatically at a set time, without you being present. Tasks allow you to automate recurring AI workflows — for example, a daily morning briefing, a weekly report generation, or an automated data-processing pipeline.Tasks are configured and managed at
/tasks.Related: Tasks documentationTool Pill
Tool Pill
A toggle button in the chat prompt bar for activating specific tools before sending a message. Tool pills appear as small labeled buttons at the bottom of the chat input area.Available tool pills:
- Search — enables web search with the message
- Deep Research — triggers a multi-step research report
- Canvas — outputs the response in the Canvas editor
- Agent — activates Agent Mode for autonomous multi-step actions
- Voice — starts a voice chat session
VAD (Voice Activity Detection)
VAD (Voice Activity Detection)
Server-side technology that automatically detects when the user has finished speaking during a voice session, so ZeroTwo knows when to start processing and generating a response.VAD eliminates the need to press a “stop talking” button — you simply pause, and ZeroTwo detects the end of your speech after a brief moment of silence (typically 1–2 seconds).If ZeroTwo seems unresponsive after you speak, wait 2–3 seconds for the VAD to register the end of your speech before assuming there’s an issue.Related: Voice Issues
Workspace
Workspace
The Business plan’s shared organizational environment with admin panel access. A workspace includes:
- A shared namespace for all Business plan members
- Admin panel at
/admin/*for managing members, billing, and settings - Member roles (Owner, Admin, Member)
- Shared connectors and org-level settings

