YourGPT

Authenticated API functions read, create, update or delete records through HTTP requests, with headers for credentials and typed parameters marked as required when needed. Builders can import a cURL definition or implement a function in JavaScript or Python. The agent fills the inputs from the conversation and uses the response in its next step. Connected functions can change business data, while remote MCP servers supply additional tools.

Sourcedocs.yourgpt.ai/chatbot/ai-agent-management/functionsread 2026-10-20

AI Studio combines conditions, API calls, code steps and autonomous agent nodes in an editable flow. Customers route requests among specialist agents and use variables to pass results into later steps. JavaScript or Python handles transformations that need code, and webhooks start flows from external events. Named versions can be published, compared, exported or restored, with an emulator for trying a changed workflow before releasing it.

Sourceyourgpt.ai/ai-studioread 2026-10-20

Customer knowledge comes from website pages, uploaded files, FAQs and connected content services such as Google Drive, Notion and Confluence. Source refresh frequency depends on the subscription. Smart Learning identifies unresolved questions and suggestions from earlier chats; staff review the proposed FAQ answers before adding them to the collection. Corrections improve the material retrieved for later answers, while conflicting source content can be flagged for an owner.

Sourceyourgpt.airead 2026-10-20

Check with Team asks a teammate for information or a decision without transferring the entire conversation. In a dashboard card, the teammate enters an answer, generates a polished draft and edits it before delivery. The response stays held until that person selects Approve & send. Only members with conversation editing access can reply, and notifications can reach Slack, Discord or a webhook while the agent waits for the decision.

Sourcedocs.yourgpt.ai/chatbot/ai-agent-management/functionsread 2026-10-20

Organization roles distinguish ownership, editing, viewing and billing responsibilities, while project roles include operators and trainers with different access, and billing access does not grant access to project content. Enterprise identity integration supports SAML 2.0 and OAuth through Custom SSO. YourGPT holds SOC 2 Type II and ISO 27001 and states GDPR compliance. Encryption protects stored data and connections, customer conversations are excluded from model training, and data isolation separates each customer's workspace.

Sourceyourgpt.ai/enterpriseread 2026-10-20

Session Logs open from the right sidebar of an individual conversation and show what the running workflow executed in production. Each entry carries a timestamp, the conversation identifier and an event type such as node trigger, node ending or flow completion, and operators can expand the structured JSON data or pick out debug and error events next to the messages received, which shows the node where a flow failed or stopped. The same debugging area holds audit logs, a training audit and conversation logs for reviewing what the agent answered.

Sourcedocs.yourgpt.ai/chatbot/debugging/conversation-logs/session-logsread 2026-10-20

Flow variables hold values collected or calculated within a workflow, and session variables track conversation state and operating mode. Visitor and contact fields carry details such as language, email, phone and tags. Builders reference these values in later nodes, while the agent's previous-message limit controls the conversation context sent to the model. Customer records and reviewed FAQ updates persist separately from the session's working context.

Sourcedocs.yourgpt.ai/chatbot/studio/elements/variablesread 2026-10-20

Standard subscriptions run on the managed YourGPT service, and no setting lets a customer pick a hosting region. Enterprise contracts offer custom agent deployment arrangements and retention settings, with the order itself defining the infrastructure, but no page names a private cloud, on premises or regional option. The data processing agreement requires consent before EU and EEA personal data is transferred, with an exception for essential operations, and model providers run with zero day retention so prompts are not stored on their side.

Sourceyourgpt.ai/pricingread 2026-10-20

AI Studio provides importable starting flows for named business tasks, including Zendesk ticketing, Salesforce lead synchronization, Stripe integration and a handoff to Crisp. Freshdesk ticketing and Zapier lead synchronization are additional options. Customers import a template, add their own credentials and test the connected steps before use. The template library also includes commerce starting points for Shopify, WooCommerce and BigCommerce, which customers adapt to their own store.

Sourceyourgpt.ai/ai-studioread 2026-10-20

Incoming conversations start the agent through a website widget or connected services such as WhatsApp, Instagram, Telegram, Slack and phone. AI Studio can also react to external webhook events, including a payment event from Stripe, and route the resulting work through the configured flow. The same workspace handles chat and voice channels, with handover to staff when the conversation reaches a condition requiring their involvement.

Sourceyourgpt.ai/ai-studioread 2026-10-20

Customers select among models from OpenAI, Anthropic, Google, Meta, Mistral, xAI and other providers for each agent. The model table marks which choices support functions and AI Studio, so an agent that needs tools narrows the list, and higher plans unlock stronger models such as o1 on Advanced. Credit consumption changes with the selected model, prompt size, response length and any added processing. Enterprise customers can also connect custom models of their own.

Sourcedocs.yourgpt.ai/chatbot/other/ai-and-usage/ai-modelsread 2026-10-20

The REST interface has project keys for a single agent project and platform tokens for organization scope. Project endpoints cover conversations, voice and knowledge operations; platform endpoints extend to organizations, projects, training data, contacts and analytics. Requests use authenticated JSON calls. Project API access starts on Professional, while the broader platform token is available to Advanced, Agency and Enterprise customers. A widget SDK embeds the conversation interface in applications.

Sourcedocs.yourgpt.ai/chatbot/api-referenceread 2026-10-20

Agent Playground loads the agent's persona, restrictions, knowledge and tools, then runs one test input across up to five model emulators at once. Reviewers compare the answers side by side with response time, token use, credits and cost, and saved presets keep test settings such as the previous message limit and temperature for the next run. API, system, code and MCP functions and Studio intents all execute in the playground, so a change to a tool is tested with the agent that will call it. Results are compared by a person; there is no automatic score.

Sourcedocs.yourgpt.ai/chatbot/debugging/agent-playgroundread 2026-10-20

AI Copilot runs inside the customer's own website through the widget SDK once Agent Mode is enabled. Built in actions switched on with a toggle and no code let it work the visitor's live page: Navigate moves to another page or section, DOM Manipulation clicks controls, fills forms and interacts with page elements, Capture Screenshot sees what the visitor sees, and Console Logs and Network Requests inspect errors when something breaks. The agent decides when to use each action. Its reach ends at the site where the widget is installed, with no control of other applications or the desktop.

Sourcedocs.yourgpt.ai/chatbot/developer-guide/widget/ai-copilotread 2026-10-20