--- title: "Dialogflow Integration" description: "Connect AI agents to Google Cloud Dialogflow ES. Agents detect intent from natural language messages, fire events to start scripted conversations, manage intents, entity types, and session contexts, and retrain the agent programmatically." url: https://flowrunner.ai/integrations/google-cloud-dialogflow date_modified: 2026-08-07T17:20:23-07:00 --- # Dialogflow [AI](https://flowrunner.ai/integrations/category/ai-llms) Connect AI agents to Google Cloud Dialogflow ES. Agents detect intent from natural language messages, fire events to start scripted conversations, manage intents, entity types, and session contexts, and retrain the agent programmatically. [Verified](https://flowrunner.ai/integrations/verified "What does verified mean?") · 20 actions · OAuth · available [Dialogflow website](https://cloud.google.com/) · [Platform Documentation](https://cloud.google.com/dialogflow/es/docs) · Capability data verified 2026-08-12 1. A customer message arrives from the website chat 2. Detect Intent classifies the message and extracts its parameters 3. The agent inspects match confidence and the fallback flag before acting 4. A support ticket opens with the extracted parameters attached 5. The fulfillment reply goes back to the customer in the chat 6. Low-confidence conversations hand off to a person with the full transcript ## What This Integration Enables Dialogflow ES answers one question well: what does this customer want? It does not finish the work. That gap is where most chatbot projects quietly die, a bot that classifies requests perfectly and then leaves them in a log. This connector closes the loop. FlowRunner agents send messages through Detect Intent, read the matched intent and its extracted parameters, and turn the answer into completed work in the systems behind the conversation: a ticket, a spreadsheet row, a follow-up email. The connector also covers the maintenance side that conversational teams usually do by hand in the console: creating and updating intents and entity types, steering live conversations through session contexts, and retraining the model with Train Agent. That last group changes what a production bot says to real customers, which is why FlowRunner puts [human-in-the-loop](https://flowrunner.ai/concepts/human-in-the-loop) review between staged changes and the retrain. ### Without FlowRunner **The chatbot cul-de-sac**: The bot understands the request and then nothing happens **Blind retraining**: Intent changes ship straight to the production model without review **Escalation resets to zero**: Handoffs make the customer repeat everything to a human ### With FlowRunner **Intents become actions**: A matched intent opens tickets, logs leads, and sends follow-ups **Reviewed model changes**: Intent edits are staged and a person approves before Train Agent runs **Context travels with the handoff**: The human picks up with parameters and transcript in hand ## Use Case Scenarios ### The intent that becomes a ticket A customer types a support request into the website chat. Detect Intent matches it and extracts the parameters: product, issue type, order number. The agent checks confidence, then opens a ticket in [Zendesk](https://flowrunner.ai/integrations/zendesk) populated with the extracted fields and the customer's own words, and the fulfillment reply confirms the ticket number in the chat. If the match is weak or the fallback intent fired, the agent does not guess at the customer's problem. It hands the conversation to a person with the transcript and every extracted parameter attached, so the customer never repeats themselves. ### Lead capture that lands in the pipeline A lead-capture flow walks a visitor through qualifying questions. When Detect Intent resolves the final step, the agent writes the collected contact details and the detected intent to [Google Sheets](https://flowrunner.ai/integrations/google-sheets) for the sales team, and Detect Intent by Event fires a follow-up event that closes the conversation cleanly. A summary email goes out through [Gmail](https://flowrunner.ai/integrations/gmail-service) with next steps. The conversation the bot had becomes a lead the team can actually work. ### The agent that maintains the agent The conversational team collects new phrasings customers actually use. A FlowRunner agent stages them: Create Intent for genuinely new requests, Update Intent to add training phrases to existing ones, Create Entity Type when a new product line needs recognizing. The staged changes post for review, and only after a person approves does Train Agent run. Session contexts get the same treatment in live operations, with Create Context steering conversations and Delete All Contexts resetting a stuck session for a fresh start. ## Human-in-Loop Highlight Delete Intent is permanent, and Train Agent is what makes it real: once the model retrains, the live agent immediately stops recognizing those requests, for every customer, in production. That pairing is exactly where an autonomous loop should stop. When a cleanup agent identifies intents that look retired, old campaign flows, deprecated product names, it stages the removals and asks the conversation owner in Slack: "Proposing deletion of 6 intents including promo-2025-signup and legacy-return-flow, then a retrain. These cannot be restored after deletion. Approve the list, edit it, or hold?" Only after approval does the agent call Delete Intent on the confirmed list and then Train Agent to make the change live. A model your customers are talking to right now does not get edited unattended. Agent processes routinely Detects exception requiring judgment Clear match Continues automatically Ambiguous Routes to human via Slack Human decides Agent resumes with decision ## Agent Capabilities 20 actions ### Conversation 2 - **Detect Intent** Sends an end-user text message to the Dialogflow ES agent and returns the detected intent with its fulfillment response. The core action for chatbot and assistant flows. - **Detect Intent by Event** Triggers a named event in a session instead of text, returning the matched intent and fulfillment. Used to start scripted flows like a welcome greeting or a follow-up sequence. ### Agent Management 3 - **Get Agent** Retrieves the agent's configuration including display name, default language, and time zone. Used to confirm which agent a flow is talking to. - **Search Agents** Searches the GCP project for available agents and returns their configurations. Used for discovery across projects. - **Train Agent** Retrains the agent's model so intent and entity changes take effect in production. Long-running, consequential, and always run after human-approved changes. ### Intents 5 - **List Intents** Lists the agent's intents with pagination. Used to audit the conversational surface. - **Get Intent** Retrieves an intent with its training phrases, priority, and contexts. Used to load current state before proposing edits. - **Create Intent** Creates an intent from a display name, training phrases, and responses. Used to teach the agent new requests programmatically. - **Update Intent** Changes only the supplied fields on an intent. Used to add training phrases and refine responses without touching the rest. - **Delete Intent** Permanently deletes an intent. Cannot be undone, and takes effect at the next retrain; always behind human review. ### Entity Types 5 - **List Entity Types** Lists the agent's entity types with pagination. Used to audit what the agent can extract. - **Get Entity Type** Retrieves an entity type with its entities and matching settings. Used before extending vocabularies. - **Create Entity Type** Creates an entity type as a synonym map or list. Used when new products or categories need recognizing. - **Update Entity Type** Changes supplied fields on an entity type. Used to grow vocabularies as the catalog changes. - **Delete Entity Type** Permanently deletes an entity type. Cannot be undone; gated with the same review as intent deletions. ### Session Contexts 5 - **List Contexts** Lists the active contexts of a conversation session. Used to inspect where a conversation stands. - **Get Context** Retrieves a context with its remaining lifespan and parameters. Used to debug a specific conversation state. - **Create Context** Activates a context in a session, steering which intents can match next. Used to guide conversations from outside the bot. - **Delete Context** Deactivates a single context without touching others. Used for targeted state corrections. - **Delete All Contexts** Clears every active context in a session. Used to reset a stuck conversation for a clean start. ## Frequently Asked Questions ### What can FlowRunner do with Dialogflow? FlowRunner agents can run Detect Intent, Detect Intent by Event, and Get Agent in Dialogflow, plus 17 more actions. ### Does connecting Dialogflow to FlowRunner require OAuth? Yes. Dialogflow connects to FlowRunner with OAuth 2.0, so agents authenticate without handling raw credentials. ### Can Dialogflow trigger a FlowRunner workflow automatically? Dialogflow doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger. **Work at Dialogflow?** This integration exposes Dialogflow to AI agents on every FlowRunner plan, including through MCP, at no cost to you. [See what FlowRunner offers integration partners](https://flowrunner.ai/integrations/partners), including how to keep this page current. --- Markdown version of https://flowrunner.ai/integrations/google-cloud-dialogflow. Site index: https://flowrunner.ai/llms.txt