YourGPT Chatbot
AIBuild, train, and operate no-code AI support agents with the YourGPT Chatbot API and drive conversations through it. Agents keep support answers grounded in current documentation.
What This Integration Enables
YourGPT Chatbot is the connector in this category that measures itself. Get AI Resolution Statistics returns how many sessions the AI resolved on its own against how many were escalated to a human operator, over a date window. Get Top Intents returns the visitor intents that actually came up, with counts, which is a coverage backlog handed to you rather than assembled from complaints. List Execution Logs shows the flow actions and runtime events behind an answer, so "why did it say that" is a query rather than an argument. Put those three together in a FlowRunner flow and you get a closed loop: what were people asking, did the bot handle it, and if not, what is missing.
The project is also a metered object with real ceilings on it. Get Usage Limits reports both the configured ceiling and current consumption for documents, web pages and message credits, so a flow can check headroom before it starts a batch of training. Set Usage Limits applies caps, and Update Project Settings controls the system prompt, restrictions, model, temperature, how much history the bot carries, how many knowledge nodes it may consult, and how long an idle conversation waits before it auto-closes. Alongside that sits a small contact layer with lists, tags and companies, an operator handover through Update Session, and a Make Outbound Call action that hands a phone number to an AI Voice Agent. It is closer to an instrumented support desk than to a bot builder, and it rewards being run as one.
Without FlowRunner
With FlowRunner
Use Case Scenarios
A deflection number that turns into a knowledge backlog
Every Monday the flow pulls Get AI Resolution Statistics for the previous week and Get Top Intents alongside it. The two together answer a question a satisfaction score never does: which specific topics are driving escalations. For each frequent intent the agent runs Query Knowledge Base to see what the corpus would actually surface, and pulls List Execution Logs on one sample session so it can distinguish "we have nothing on this" from "we have it but the answer never reached the customer". Confirmed gaps are drafted as Q&A entries and staged. The support lead reads a ranked list with counts attached and approves the ones worth writing. Nothing goes into the knowledge base without that approval, because a bad entry is a wrong answer repeated at scale.
A session handed to an operator, then closed by the workflow
A conversation goes past what the bot should handle. The agent calls Update Session to switch the chat mode to Operator, which takes the AI out of the thread, and posts the transcript from List Session Messages into Slack with the visitor's contact record from Get Contact Detail attached. When the operator resolves it, the flow calls Update Session again to mark the resolution state and applies a segment label so the case is findable later. On New Contact runs in parallel, restricted to the leads segment, syncing genuinely new contacts into HubSpot rather than every visitor the widget ever saw.
A voice follow up that a person authorises
Make Outbound Call hands a number in E.164 format to an AI Voice Agent and puts it through, with extra data passed along so the call carries context from the workflow. That is a real phone ringing on a real person's desk, and it cannot be recalled once it starts. So the flow assembles the case rather than the call: it identifies the chat sessions where a callback was requested, resolves each contact with Get Contact Detail, drafts the context payload the Voice Agent will use, and posts the list to the owner with the reason each person is on it. The owner approves the batch or trims it. The agent gathers, the person authorises, and only then does anything dial.
Human-in-Loop Highlight
The gate is Update Project, and specifically the retention field, because it is the operation most likely to be mistaken for a settings change when it is actually a standing deletion policy.
Retention on a YourGPT project is expressed in seconds, and the documented reference values are 0 for keep forever, 2592000 for 30 days and 7776000 for 90 days. Moving that number down does not delete anything visible today. It commits the project to deleting conversation data continuously from now on, and everything that ages past the new boundary goes without a per record notice and without a way back. A workflow tidying up project configuration can write that field in the same call it uses to rename a project, and the response is only a status envelope, so nothing about the result tells you that a retention policy just changed. Months later someone goes looking for the conversation behind a dispute and it is not there.
So a flow that touches project configuration splits the write. Renames and cosmetic changes it makes on its own. A retention change it stages and asks about, in the terms the decision is actually made in: "Update to the EU support project would change retention from keep forever to 30 days. The project currently holds conversations back to March. Applying this deletes everything older than 30 days on a rolling basis, permanently, and no export exists. Confirm, or leave retention unchanged and apply the rename only?" A person owns that, because retention is a legal and commercial decision wearing the clothes of a configuration field. Set Usage Limits gets a lighter version of the same treatment for a structural reason: the call replaces the limits currently configured rather than merging with them, so a flow that sets one cap silently drops the others, and the first sign of trouble is a bot that stops answering customers mid afternoon.
Agent Capabilities
30 actionsProjects and Limits
6- List Projects Returns the chatbot projects the API token can reach, each with its numeric id, project UID, organization, member roles and widget UID. The project UID returned here is the identifier every other operation expects.
- Create Project Creates a new chatbot project in the organization tied to the token and returns the generated project UID and widget UID. Only a display name is required; everything else is configured afterwards.
- Update Project Renames a project and changes how long its conversation data is retained. Retention is expressed in seconds, with documented reference values for keep forever, 30 days and 90 days. The response is only a status envelope, not the updated project.
- Set Usage Limits Applies one or more consumption caps to a project. Each entry pairs a limit type with a ceiling, and the call replaces the limits currently configured rather than merging with them.
- Update Project Settings Configures how a project answers: system prompt, restrictions, language model, sampling temperature, how much chat history it carries and how many knowledge base nodes it may consult. Also controls whether source page links are appended to answers and how long an idle conversation waits before it auto-closes.
- Get Usage Limits Returns the configured ceiling and current consumption for each metered resource, covering documents, web pages and message credits. Checked before a workflow starts a batch of training or conversations.
Analytics
4- Get Statistics Returns aggregate performance figures for a project including total conversations, messages, documents and links, visitor and session counts, thumbs up and thumbs down feedback, and average messages per conversation, over an optional date window.
- Get AI Resolution Statistics Returns how many sessions the AI resolved on its own against how many were escalated to a human operator, alongside the total session count. The basis for tracking deflection over a window and alerting when escalations rise.
- Get Top Intents Returns the most frequent visitor intents detected on a project with a count for each, so recurring questions can be turned into new knowledge base entries.
- List Execution Logs Returns chatbot execution logs showing the flow actions and runtime events behind answers, which is the fastest way to establish why an agent replied the way it did. Narrowable to a single session and a date range.
Conversations
5- Create Session Opens a new chat session against a project and returns its session UID, chat mode, state and visitor metadata. A session is the container every message belongs to, so this comes first when driving a conversation from your own channel.
- Send Message Posts a visitor message into an existing session and returns the AI agent's reply with its message id and content type. Used to run a conversation from a surface other than the widget.
- List Sessions Returns chat sessions for a project with status, chat mode, state, country, last message and last message time, plus visitor, contact and session metadata. Results can be narrowed to open or closed sessions and to a date range.
- List Session Messages Returns the transcript of a single session including who sent each message, its content type, any answer choices, attachment URL, response source and rating. Used to archive a conversation or feed a summary step.
- Update Session Changes the resolution state, handling mode or segment label of a session. This is how a workflow marks a conversation resolved, or hands it from the bot to a person by switching the chat mode to Operator.
Knowledge Base
8- Query Knowledge Base Runs semantic search across a project's trained knowledge and returns the best matching content nodes. Used to ground an external prompt in the same material the bot answers from, or to check what a question would retrieve before it is asked.
- Add Training Text Adds a text or FAQ entry to the knowledge base. FAQ stores a question and answer pair, with the short text holding the question and the detail holding the answer; Text stores a standalone passage.
- Add Training URL Queues a web page for crawling into the knowledge base and returns the created source with its pending status. The page can optionally be scheduled to re-crawl weekly or monthly so the bot keeps answering from current content.
- List Training Files Returns the documents uploaded to a project's knowledge base with name, indexing status, byte size and tags. Used to confirm an upload finished indexing before relying on it.
- Get Training Status Reports whether a project's knowledge base is still indexing or has finished. Polled after adding text, URLs or files to hold a workflow until the bot is ready to answer from the new material.
- Create Knowledge Base Tag Creates a tag that can be attached to knowledge base entries, files and URLs so training material can be grouped and filtered. An existing tag with the same name is returned rather than an error, which makes it safe to call repeatedly.
- Upload Training File Adds a document to the knowledge base by pointing YourGPT at a publicly reachable file URL, which it downloads and indexes. Indexing is asynchronous, so Get Training Status is polled before the content is relied on.
- Link Knowledge Base Tags Assigns a set of tags to a single knowledge base asset by its id and kind. The supplied tags replace every tag currently on that asset, so existing tags must be included when adding to them.
Contacts
6- Create Contact Lead Creates a lead contact on a project so a person captured outside the widget is tracked alongside chat generated contacts. Either an email address or a phone number must be supplied.
- Get Contact Detail Returns the full record of a single contact including identity fields, segment type, tags and linked company. Used after List Contacts or the On New Contact trigger to enrich a workflow.
- List Contacts Returns the people captured by a project with name, email, phone, segment type, tags and company. Narrowable by free text search, segment, tags, contact list and creation date range, with spam excluded.
- List Contact Lists Returns the contact lists defined on a project, each with its filter definition and the number of contacts currently matching it. The returned list id narrows List Contacts to a single audience.
- List Contact Tags Returns the tags available for labelling contacts, with a total count for pagination. Used to discover the tag vocabulary before filtering contacts.
- List Contact Companies Returns the companies associated with a project's contacts, each with its website and contact count. Used to roll conversation activity up to an account before syncing to a CRM.
Voice
1- Make Outbound Call Triggers an AI Voice Agent to place an outbound phone call and returns the call id and its initial status. The destination number must be in E.164 format, and extra data is passed through so the call can be personalised with context from the workflow.
Triggers
2 triggersEvent Triggers
2- On New Session Fires when a visitor starts a new chat session on the selected project, emitting the full session record including status, chat mode, state, country, last message, and visitor and contact data. YourGPT exposes no webhook registration API, so this trigger polls the sessions endpoint rather than writing to the project's webhook configuration, leaving dashboard settings untouched. The first poll establishes a baseline without firing.
- On New Contact Fires when the selected project captures a new contact, emitting the record with name, email, phone, segment type, tags and company. Contacts can be restricted to a segment such as leads only. This trigger polls the contacts endpoint, and the first poll establishes a baseline without firing.
Frequently Asked Questions
What can FlowRunner do with YourGPT Chatbot?
FlowRunner agents can run List Projects, Create Project, and Update Project in YourGPT Chatbot, plus 27 more actions.
Does connecting YourGPT Chatbot to FlowRunner require OAuth?
No. YourGPT Chatbot connects to FlowRunner with an API key, no OAuth flow required.
Can YourGPT Chatbot trigger a FlowRunner workflow automatically?
Yes. YourGPT Chatbot supports 2 triggers that can start a FlowRunner workflow automatically.
Start building with YourGPT Chatbot
$100 in credits. No card required. Connect in minutes.