Custom GPT Agents
Last updated
Last updated
This documentation explains custom GPT agents, how to set them up in Bind AI Copilot.
Custom GPT agents operate on your instructions provided via "System Prompt" and can perform the specified actions. You can also upload/attach data which the GPT can specifically use. Custom GPTs can understand your preferences and work requirements and deliver desired outputs. You can also automate repetitive tasks to consistent outputs and save time in both business and personal projects, or restrict the type of tasks the GPT can perform (e.g. GPT for Language Translation).
Go to Bind AI Copilot's interface and click that 'robot' icon at the bottom of the right menu.
Name Your GPT Agent: Pick a name that's clear and descriptive.
Provide a Description: Write a brief overview explaining what your agent does. A sentence or two should be enough.
Start by thinking and theorizing about your needs, their challenges, and what capabilities your agent should have to perform optimally. A clear purpose and sysurltem prompt is key to building an effective agent.
Use the 'RACE' model to outline your agent's instructions:
Role: Describe your GPT agent's role (e.g., "You are a customer support assistant").
Actions: Specify its tasks (e.g., “Answer customer inquiries”).
Context: Provide any relevant background or data. You can also use upload file or connect integrations which will automatically retrieve the relevant context.
Examples: Include sample responses to guide expected outputs.
If your agent needs specific information, upload documents or data to help it give accurate responses. Bind AI also provides native GitHub and Google Drive support. Read this guide to know how it works.
Once your create your Agent, it will show the unique URL for the agent.
You can find in the list of bots, which are in the top section (next to Web Search, Code Generator)
Once set up, test your custom GPT agent in real scenarios to see how it performs and adjust as needed.
A: No. Custom GPT agent features are only available on paid plans.
A: The RACE model (Role, Actions, Context, Examples) helps structure your agent's instructions. It's a helpful guideline for defining how your agent behaves and responds.
A: Not currently. However we are working on allowing Edit and Archive functionality for created Agents.
A: Select the created agent and try some queries to see if it follows the instructions you provided.
A: You can upload any document that contains helpful information for your agent. That can include FAQs, customer guidelines, or relevant data or code to improve response accuracy. Bind AI Copilot provides GitHub and Google Drive integration, so uploading files won't be a hassle.
