How to pick the right agent, what each specialist does, and how to get the best results from a conversation.
The Agents page is the main conversational interface. Each agent has a distinct role, a set of tools, and a focused set of capabilities. Understanding what each one does helps you phrase requests more effectively and get higher-quality outputs.
When you start a conversation, your message goes through a lightweight router that reads your intent and selects the best agent for the job. You can also start from the agent selection screen and pick one directly — useful when you know exactly which specialist you need.
If an agent receives a request outside its scope, it will acknowledge the limitation and suggest the right agent rather than producing a poor-quality answer.
The front-door agent. Handles open-ended questions, quick lookups, and anything that doesn't clearly belong to a specialist. If you're not sure where to start, begin here — the assistant will guide you to the right tool or agent.
Best for: written content, social media, SEO, research-backed drafts.
The content strategist can search the web, extract information from URLs, generate structured outlines, write platform-formatted posts (LinkedIn, Twitter/X, Instagram, Newsletter), and generate supporting images. It works best when given a clear topic, target audience, and tone.
Example prompts:
Best for: short-form video content (Instagram Reels, TikTok, YouTube Shorts).
Produces a complete video package: a script formatted for on-camera delivery, a reel brief with hook/CTA structure, and a matching cover image. Provide the topic, platform, and target audience for best results.
Best for: long-form video, podcast episodes, webinar scripts.
Produces structured scripts with intro, sections, and outro. Supports different formats (YouTube explainer, interview-style podcast, keynote-style webinar). Provide the topic, desired length, and any key points to cover.
Best for: YouTube thumbnail concepts and cover images.
Generates a visual concept brief describing the composition, color scheme, and text overlay — then produces an AI image based on it. Works best when you provide the video title and target emotion (curiosity, urgency, excitement, etc.).
Best for: extracting structured data from receipt photos.
Upload a photo of a receipt and the agent extracts merchant name, date, individual line items, tax, and total into a structured JSON output. Useful for expense reporting or accounting workflows. Accuracy depends on image quality.
Best for: reviewing contracts and legal documents.
Upload a PDF or paste contract text, choose an analysis focus (general summary, risk identification, or key dates/deliverables extraction), and the agent returns a structured Markdown report. The agent can also email the analysis to a specified address.
Be specific about format. "Write a LinkedIn post" is fine, but "Write a 150-word LinkedIn post in a conversational tone, ending with a question to drive engagement" gives the agent a clear target.
Include the platform. Social copy for Instagram reads differently from LinkedIn. Mention the platform whenever relevant.
Provide source material when you have it. Paste a product description, a URL, or a document. Agents can work from your content rather than generating from scratch.
Iterate. After the first output, follow up: "Make the headline punchier" or "Shorten this to three sentences" — the agent retains the full conversation context and can refine without starting over.