Skip to main content
Every agent on your team has a set of instructions that defines what it does, how it communicates, and what processes it follows. The more context you give it, the more useful it becomes.

​
Instructions, tools, and skills

Instructions are what make your SEO Strategist think like an SEO expert and your Sales Outreach Specialist write like a seasoned salesperson. They define each agent’s approach — the questions it asks, the formats it uses, and the context it keeps in mind. Beyond instructions, each agent comes with built-in tools — capabilities like web search, file creation, and data handling that let it produce real work, not just responses. For complex, repeatable processes, agents can also have skills: specialized routines like “analyze a business proposal” or “create a content calendar.” Shape any agent’s behavior just by talking to it:
  • “Our brand voice should always be warm and approachable”
  • “When writing outreach emails, always mention our free trial”
  • “Keep blog posts under 800 words unless I say otherwise”
  • “Reference the customer’s industry when making recommendations”
Agents remember these preferences and apply them going forward. The more context you share, the better the results.

​
How Appy chooses your agents’ models

Appy runs different agents on different models, chosen to optimize output quality while minimizing cost. A research-heavy specialist and a quick-turnaround writer don’t need the same model underneath them, so we don’t give them the same one — we match each agent to the model that gets the best result for what it does. We use frontier-generation models from providers like Anthropic, OpenAI, and others, and we update those choices as better options become available. Model selection isn’t something you configure. That’s by design: matching model to task is work the platform does so you don’t have to think about it.