Chat
The chat surface is an AI-powered command center for the workspace. Instead of a blank prompt, it opens to a Tool Catalog organizing suggested prompts into four action categories plus a Smart Picks row of personalized suggestions drawn from your actual projects, tasks, and documents. Conversations can also originate from Slack and Telegram via bidirectional delivery channels.
Key Concepts
- Tool Catalog — Four category tabs (Explore, Create, Debug, Automate) plus a Smart Picks row organize suggested prompts generated from your actual workspace data, helping you discover what the assistant can do.
- @ Mentions and Context — Type
@to trigger a fuzzy-search autocomplete popover for projects, tasks, documents, profiles, and schedules. Mentioned entities are injected into the conversation context automatically. - Multi-Provider Model Selection — Choose between Claude (Haiku 4.5, Sonnet 4.6, Opus 4.6), OpenAI (GPT-4o-mini, GPT-4o via Codex), and Ollama models at $0 — with clear cost tier labels per conversation.
- Bidirectional Channel Conversations — Chat with
ainative-businessagents directly from Slack or Telegram. Channel conversations use the same engine with full tool access and permission handling, and appear in the sidebar alongside in-app chats. - Quick Access Navigation — When the assistant mentions an entity, navigation pills appear at the bottom of the response for one-click deep linking to that project, task, or workflow.
- Entity CRUD via Chat — Create, update, and delete projects, tasks, and workflows through natural language without navigating to their respective surfaces.
- Slash Command Popover — Type
/to open a tabbed popover with four views: Actions (session commands like clear, compact, export), Skills (composable behavior stacks), Tools (browser and workspace tools), and Entities (@ reference picker). - Skill Composition — Stack specialized behaviors onto a conversation from the Skills tab. Add or replace skills with conflict detection, prompt-budget eviction, and capability gating per runtime.
- Conversation Templates — Start conversations from saved blueprints via the empty-state button,
/new-from-templateslash command, or the⌘KTemplates group. - Saved Searches & Filters — Pin frequently used filter + query combinations and reuse them from the
⌘Kpalette or the@popover. Filter by namespace using#scope:,#type:, and other qualifiers. - @ Mentions with Context Injection — Type
@for a fuzzy-search autocomplete showing pinned items, projects, tasks, and documents with live status badges. Selected entities inject their full context into the conversation.
How It Works
Chat uses a progressive 5-tier context injection system that loads workspace awareness from lightweight summaries up to full document content within a ~53K token budget. This means the assistant knows your workspace state without you having to explain your setup.
Responses stream in real time with full markdown support — headings, lists, code blocks with syntax highlighting, and tables render as the text arrives. Quick Access pills appear after streaming completes, linking directly to any entity mentioned in the response.
Model Selection
Switch models at any time using the selector in the input area. Each model is labeled with its cost tier so you can balance capability against spend.
Suggested Prompts & Create Tab
The empty state displays context-aware prompt suggestions organized into tabbed categories. The Create tab surfaces entity creation options that are pre-filled with workspace context.
Quick Access Deep Linking
After asking about a workspace entity, Quick Access pills at the bottom of the response let you jump directly to the relevant detail page in one click.
Slash Commands & Composer
Type / in the composer to open a tabbed popover organizing every in-chat action. The Actions tab lists session commands — start a new conversation, compact history, export to a document, or create a task without leaving the chat. A #key:value filter tip lets you narrow results across any tab.
Skill Composition
Switch to the Skills tab to layer specialized behaviors onto the current conversation. Each skill card shows its health status, scope tags, and an + Add button. Active skills display a badge with a remove control, and a counter reads “N of M active” so you always know your stack against the runtime’s cap.
Tool Browser
The Tools tab exposes every tool available to the agent — browser automation, workspace operations, and MCP server tools. Each entry shows its parameter signature so you know what inputs are accepted before invoking.
@ Mentions & Entity Picker
Type @ to trigger a fuzzy-search autocomplete that groups results into Pinned, Projects, Tasks, and Documents. Each result shows a live status badge — completed, active, running — so you can quickly gauge what you’re referencing. You can also reach this from the Entities tab in the slash popover.
Use Cases
Daily Standup Triage
Product Team · Getting a quick status overview before the morning standup
Ask “What’s the current status of the Q2 launch project?” and get a summary of task distribution, recent completions, and blockers — then click the Quick Access pill to jump straight to the project detail view.
Task Creation from Context
Developer · Creating structured work from a natural language description
Describe what needs to happen — “Create a task to analyze the Magnificent 7 tech companies” — and the assistant creates it with proper priority, status, and description, ready for agent execution.
Workspace Discovery
New User · Learning what ainative-business can do through conversation
Browse the suggested prompt categories to discover capabilities you haven’t explored yet. The Explore tab surfaces questions about your existing data, while the Create tab helps you build new entities.
Related Features
- Task Execution — Where tasks created from chat land for agent execution
- Projects — Project context that powers chat awareness
- Workflows — Multi-step orchestration you can create through chat
- Cost & Usage — Model selection in chat directly impacts spend visibility