AI Chat

Use the AI assistant with multi-round conversations and tool calling.

7 min read

AI Chat

Talk to an AI assistant that understands your project, can query your data, create entities, trigger workflows, and push content to your CMS — all from a single chat interface. Think of it as a command center where natural language replaces clicking through menus.

What the Chat Can Do

Query Your Data

Search entities, check keyword rankings, get content statistics, and pull analytics summaries.

Create & Update

Create entities, update field values, generate content, and approve pending items without leaving the chat.

Run Workflows

Trigger publishing pipelines, bulk generation jobs, and content quality evaluations from a single message.

Getting Started

1

Open a chat session

Click Chat in the sidebar to open the chat panel. You can start a new session or continue a previous one. Each session maintains full conversation history, so the AI remembers what you discussed earlier.

2

Choose your model

When creating a session, you can select which AI model to use. Your project's default model applies if you do not choose one. You can change the model mid-session via session settings.

3

Ask anything about your project

Type your message and press Enter. The AI will use its available tools to answer questions, look up data, or take actions.

Tool Calling

The AI does not just generate text — it executes real actions through a tool system. When you ask "how many blog posts are in draft status?", the AI calls the search_entities tool, queries your database, and returns a precise answer.

Core Tools

Every project comes with these built-in tool categories:

5 of 5
Category
Tools
What They Do
Entitiessearch_entities, list_entity_types, get_entity, create_entity, update_entityFull CRUD on your data. Search by title, browse schemas, create records, update fields.
Contentgenerate_content, get_generation_stats, approve_generated_content, evaluate_content_qualityTrigger AI content generation, check pending items, approve content, and run quality evaluations.
Workflowslist_workflows, run_workflow, get_workflow_run_statusDiscover available workflows, trigger runs, and monitor progress.
Marketinglinkedin_generate_post, linkedin_bulk_generate, linkedin_iterate_postGenerate LinkedIn posts, bulk-create social content, and iterate on drafts with feedback.
Projectget_project_overviewHigh-level summary of entity types, record counts, and project health.

Extension & Integration Tools

Extensions and integration connectors add their own tools. When you install an extension or activate an integration, its tools automatically become available in chat.

Tip
Install extensions to unlock more chat tools. Each extension adds its own capabilities — for example, the SEO extension adds tools for keyword tracking and search performance analysis.

Real-World Examples

Here are prompts that show the chat's power beyond simple Q&A:

Content Operations

Generate 15 glossary terms for these keywords: headless CMS, composable
commerce, API-first, content mesh, digital experience platform, MACH
architecture, omnichannel, content federation, structured content,
content modeling, content ops, editorial workflow, content governance,
personalization engine, content graph

The AI calls generate_content with the glossary terms entity type, creates a background job, and returns a link to the review page.

Data Analysis

List all blog posts that are still in Draft status. For each one, tell
me who created it and when. Then summarize: how many drafts do we have
and what's the average age?

The AI searches entities, cross-references creation dates, and gives you a structured summary with actionable recommendations.

Workflow Automation

Run the "Publish to Sanity" workflow for all approved comparison pages.
Then check the workflow status and let me know when it's done.

The AI triggers the workflow, monitors the job, and reports back with results.

Content Quality

Evaluate the quality of our latest 3 blog posts. Tell me which ones
need the most work and what specifically should be improved.

The AI runs quality evaluations scoring readability, brand voice, SEO optimization, specificity, and completeness.

Image Attachments (Multimodal)

You can attach images to your chat messages for AI vision analysis. All supported models (Gemini, GPT-4, Claude) can analyze images you upload.

How to Attach Images

There are three ways to add images to a message:

  • Click the Attach button (paperclip icon) in the chat toolbar to browse and select image files
  • Drag and drop images directly onto the chat input area
  • Paste screenshots or copied images from your clipboard (Cmd/Ctrl+V)

Supported formats: JPEG, PNG, WebP, GIF, PDF, Word (.docx), Excel (.xlsx/.xls), and text files (.txt, .md, .csv, .json). Maximum file size: 20MB. Up to 10 attachments per message.

Documents (PDFs, Word, Excel, text files) are automatically parsed — text is extracted and included in the AI context so the model can read and analyze your documents.

What You Can Do With Images

Here's a screenshot of our competitor's landing page. Analyze the layout,
messaging, and call-to-action placement. What can we improve on our own page?
Look at this chart from our analytics dashboard. What trends do you notice
and what actions should we take?

Tip
Images are stored in your project and remain part of the conversation history. The AI can reference previously shared images in follow-up messages.

Session Management

Chat sessions are saved per project and per user. You can:

  • Rename sessions to describe what you discussed (e.g., "Q1 Content Planning")
  • Archive old sessions to keep your sidebar clean
  • Delete sessions you no longer need
  • Switch models mid-session if you want a different AI provider's perspective
  • Set a system prompt to customize the AI's behavior for a specific session
  • Attach notes as context so the AI can reference your project notes
  • Attach files — images for AI vision, documents (PDF, Word, Excel) for text analysis
  • Filter tool categories to enable or disable specific tool groups per session

Note
Sessions are private to your user account. Other team members cannot see your chat history.

Configuring the Default Model

The default AI model is set at the project level in Settings > AI. This applies to new chat sessions when no model is explicitly selected. Supported model families:

3 of 3
Provider
Example Models
Best For
Googlegemini-flash-latest, gemini-2.5-pro-previewFast responses, good balance of speed and quality
OpenAIgpt-4o, o4-miniStrong reasoning, broad knowledge
Anthropicclaude-sonnet-4-20250514Nuanced writing, careful analysis
Was this helpful?

Command Palette

Search for a command to run...