Dialogflow CX Pricing (2026): Plans, Costs and Fees

Dialogflow CX pricing starts at $0.007 per chat request and $0.001 per voice second, and Google now bills it as Conversational Agents (Flows). One design choice can double the rate on every second of a call. I mapped Google's rate card and billing rules to the monthly cost of each agent design.

TL;DR: Dialogflow CX pricing

• Flows, the agent type formerly sold as Dialogflow CX, costs $0.007 per chat request and $0.001 per voice second. Playbooks cost $0.012 and $0.002.
• New users get $600 in Flows credit and $1,000 in Playbooks credit, and both activate on first use and expire after 12 months.
• An agent that starts in a Playbook bills every turn at the Playbooks rate, even on turns a Flow handles.
• CX Agent Studio charges $0.50 per session, which costs more than Flows at every chat length and call length.

Dialogflow CX pricing at a glance

Dialogflow CX pricing runs $0.007 per chat request and $0.06 per voice minute for Flows agents. Generative Playbooks agents cost $0.012 per request and $0.12 per minute.

Pricing verified against Google Cloud's Conversational Agents pricing and CX Agent Studio pages on 25 September 2026. Verify with Google before you commit.

Dialogflow CX and Conversational Agents (Flows) are the same product on the same bill. Google's console documentation confirms the Conversational Agents console replaced the Dialogflow CX console, with pricing unchanged for agents built there.

Dialogflow CX pricing by plan

Flows agent pricing at $0.007 per chat request

What's included: Deterministic agents built from flows, pages, and intents using Google's NLU. Voice runs $0.001 per second.

Google's editions page lists speech recognition, speech synthesis, and telephony inside that audio rate. The CX agent type also ships built-in test cases, redaction, and DTMF input.

Best for: Regulated call flows where every branch has to be predictable, like appointment scheduling, payments, and account verification.

Pros:

✅ Lowest unit rate on the card, at $0.06 per voice minute with speech and telephony bundled in

✅ Built-in test cases, redaction, and DTMF input come with the agent type

✅ Up to 1,000 agents per Google Cloud project

Cons:

❌ Every route and page is designed by hand, so unmatched questions need more intents or a generative fallback billed at the Playbooks rate

❌ The only free usage is the one-time $600 credit, while Dialogflow ES keeps a no-charge Trial Edition

Playbooks agent pricing at $0.012 per chat request

What's included: Generative agents written as natural-language instructions, plus data stores, generators, and generative fallback. Voice runs $0.002 per second, or $0.12 per minute.

Any request that touches a Playbook, data store, generator, or generative fallback bills as a generative request. Stacking several of those features in one turn still counts as a single request.

Best for: Support agents answering open-ended questions from help-center content, where hand-writing an intent for each question doesn't scale.

Pros:

✅ Answers questions from data stores with no hand-built route for each one

✅ A Playbook and a data store used in the same turn bill as one request

✅ Larger trial credit, at $1,000

Cons:

❌ Costs 71% more per chat request than Flows and double per voice second

❌ Starting a conversation in a Playbook bills the whole conversation at this rate, including turns a Flow handles

CX Agent Studio pricing at $0.50 per session

What's included: CX Agent Studio is Google's Gemini-based builder inside Gemini Enterprise for Customer Experience. It bills per chat, voice, or mixed session.

Evaluations, AI augmentation, and monitoring come inside the $0.50 session price. Google also lists discount options at $0.45 and $0.40 per session.

A chat session resets after 50 turns or 30 minutes of inactivity. Voice sessions switch to $0.0025 per second after 300 seconds, which works out to $0.15 per minute.

Best for: Long, multimodal conversations where built-in evaluations and monitoring cover tooling you'd otherwise buy separately.

Pros:

✅ One unit per session keeps forecasting simple

✅ Evaluations, AI augmentation, and monitoring are included in the session price

Cons:

❌ The $0.50 floor applies to every session, so an eight-turn chat costs $0.50 here and $0.056 on Flows

❌ Moving existing flow-based agents onto CX Agent Studio pricing requires a sales conversation

Dialogflow CX free trial credits and expiry

Google gives new users two separate trial credits, $600 for Flows and $1,000 for Playbooks. Both activate on first use. They expire after 12 months.

Each credit covers this much usage at list price:

• $600 in Flows credit: about 85,700 chat requests or 10,000 voice minutes
• $1,000 in Playbooks credit: about 83,300 chat requests or 8,300 voice minutes

The 12-month clock starts on first use. Say you prototype in March and launch in October. Seven months of the credit window are already gone.

At production volume, the credit goes fast. A 20,000-call voice line at three minutes per call uses up $600 of Flows credit in about 3,300 calls, which is roughly five days of traffic.

Billable events in Dialogflow CX

Google charges per request for chat and per second of audio for voice, and each turn bills by the features it used. A common misunderstanding is that every query pays the Flows rate first and the generative rate on top.

Google's hybrid billing rules settle it. A turn that uses only a Flow bills as one Flow request, and a turn that reaches generative features bills as one Playbooks request.

Chat request billing per conversation turn

A request is any API call to Conversational Agents, whether you send it directly, through an integration, or from the console. Google notes that the number of requests per task varies with agent design, so count your own API calls per turn before you forecast.

Running multi-turn chatbot testing against the live endpoint gives you a real requests-per-task count. Design-time requests, like clicking through the visual builder or listing agent resources, are free.

Text inside a voice session bills as chat, and any voice-session request that skips the audio pipeline is charged at the text request rate.

Voice billing by input and output seconds

Each voice turn bills input seconds plus output seconds, rounded up to the nearest second. Input seconds cover the time the agent listens, including the time it listens for barge-in while speaking.

Output seconds cover the full generated response. Interrupted audio still bills in full, so a caller who cuts off a 12-second answer after two seconds still costs you 12 seconds.

Webhook time and API latency add no billed seconds. Google excludes non-audio processing from the audio total, so a slow webhook costs caller patience and nothing on the voice meter.

Rounding happens once per turn, on the combined listening and response time, so a 20-turn call adds at most 20 billed seconds.

Hybrid agent billing when Flows and Playbooks mix

The agent's entry point decides the rate. Google's pricing page splits hybrid agents into two cases.

• A Flow calls a Playbook: Only the turns (chat) or seconds (voice) that use generative features bill at the Playbooks rate. Google's 40-second example with 10 generative seconds bills 30 Flow seconds and 10 Playbooks seconds.
• A Playbook calls a Flow: The whole conversation bills at the Playbooks rate. That same 40-second call bills 40 Playbooks seconds.

For an agent that mostly follows scripts and uses generative fallback for stray questions, start the conversation in a Flow. You pay the higher rate only on turns that reach the fallback.

Costs outside the Dialogflow CX rate card

Several charges fall outside the per-request and per-second rates and can appear on the same Google Cloud invoice. Webhooks also run on your own compute, such as Cloud Run, and bill under that product.

Speech-to-Text and Text-to-Speech charges

Speech costs depend on who runs the audio pipeline. When Conversational Agents handles the audio, recognition and synthesis sit inside the $0.001-per-second Flows rate.

When your app handles speech itself and sends text to the agent, you pay three meters:

• Text-rate requests to Conversational Agents
• Speech-to-Text at $0.016 per minute on the V2 standard model for the first 500,000 minutes each month
• Text-to-Speech per character after each voice type's free tier, from $4 per million for Standard voices, which Google now lists as legacy, to $30 per million for its latest Chirp 3 HD voices

Illustrative example: Take a three-minute call with eight turns and 1,200 characters of agent speech. On the split stack, that's $0.056 in Flows requests, $0.048 for three streamed minutes of Speech-to-Text, and $0.036 for Chirp 3 HD audio.

That adds up to about $0.14 per call against $0.18 on native Flows voice. The 22% saving has a price. You now own endpointing, barge-in, and latency.

Data store index storage

Data stores include 10 GiB of index storage free each month, then bill $5.00 per GiB of raw data. Google samples the size regularly and averages it for the month, and refreshing the data costs nothing extra.

Google sizes each website page at 500 KiB, so a 1,000-page site counts as about 0.48 GiB. Google's $2.38 a month example applies the $5 rate before the 10 GiB free quota, so inside that quota a site that size adds nothing to the bill.

Test and QA traffic on the production bill

Test conversations bill like customer conversations. A regression suite that dials the agent or calls the API generates the same runtime requests Google charges for.

Take a 150-scenario voice suite run nightly for 30 days at three minutes per call. That's 810,000 billed seconds, or $810 a month on Flows and $1,620 on Playbooks.

The same suite as text, at eight turns per scenario, costs $252 on Flows. Cekura's chat testing docs recommend text runs for workflow and regression checks, and voice runs for ASR, TTS, and final production checks.

Dialogflow CX pricing calculator example at 20,000 calls

This worked example models a healthcare scheduling line with 20,000 voice calls a month at 180 billed seconds per call. The inputs are illustrative. I ran them at list price with no discounts.

Moving the entry point from a Flow to a Playbook doubles the bill with zero change in call length. The Flow-led design keeps generative answers for an extra $720 a month.

For chat, the same logic applies at 30,000 sessions of eight turns each, which is 240,000 requests.

CX Agent Studio beats Playbooks in a narrow band. A Playbooks voice call passes $0.50 at 250 seconds. After 300 seconds, Google bills Agent Studio voice at an overage rate of $0.0025 per second.

If that overage covers only the seconds past 300, Agent Studio stays cheaper until about 500 seconds. If it reprices the whole call, the band closes at 300. Confirm the reading with Google before you model long calls.

On chat, the band sits between about 42 and 50 turns per session, assuming one request per turn. Before you finalize a forecast, add data store storage, any separate speech APIs, and test traffic from the sections above.

How to reduce Dialogflow CX costs

1. Start every conversation in a Flow, so only generative turns pay the Playbooks rate.
2. Run regression suites as text, which costs $252 a month instead of $810 for the same 150 nightly scenarios.
3. Let Conversational Agents run the audio unless you can own endpointing and barge-in.
4. Keep data stores under the 10 GiB free index quota.
5. Count requests per turn on the live endpoint before you forecast.

Dialogflow CX vs. ES pricing comparison

Dialogflow ES costs $0.002 per text request, so the CX Flows rate is 3.5x higher per text request. ES is Google's older standard agent type, sized for small to medium agents.

ES wins on raw per-request cost for simple text bots. CX is the better buy once you need DTMF keypad input, redaction, or more than one agent per project.

On voice, ES splits the meters. ES bills caller audio at about $0.026 per minute before any synthesis or phone gateway charge, and CX folds both directions into $0.06.

If your agent replaces a keypad menu, plan for IVR testing on both the DTMF and spoken paths.

Choosing a Dialogflow CX plan by use case

Choose Flows if you:

• Run regulated workflows like scheduling, payments, or identity checks where every branch has to be auditable.
• Handle high call volume at short handle times, where $0.06 per minute keeps unit costs low.

Choose Playbooks if you:

• Answer open-ended questions from a large, changing knowledge base.
• Can route into a Playbook from a Flow, so only generative turns pay $0.012.

Choose CX Agent Studio if you:

• Run multimodal sessions, or voice calls that land between 250 and 300 seconds.
• Want Google's evaluations and monitoring bundled into one per-session price.

Dialogflow CX value assessment

Dialogflow CX earns its cost when call flows are structured, and volume is high. At $0.06 per minute with speech and telephony included, Flows gives you a full voice pipeline on a single per-second meter.

The costs that grow over time are design labor and test traffic. Every new route needs building and regression checks, and those checks bill against the same meter.

Dialogflow CX is worth it if you:

• Already run on Google Cloud and want billing, access control, and data stores in one project.
• Need deterministic control for compliance-sensitive flows, with redaction built in.

Skip Dialogflow CX if you:

• Run a small text-only bot, where ES at $0.002 per request costs less.
• Want a voice platform that bundles the LLM, voice, and telephony choices into one per-minute rate.

Dialogflow CX alternatives and pricing comparison

Each alternative meters a different unit. Compare the one that matches your traffic.

The Lex figures are request-and-response rates. Streaming conversations bill separately. Retell builds its per-minute range from infrastructure, LLM, voice, and telephony components. Its $0.07 floor sits close to the Flows voice rate with a model included.

How do you test and monitor a Dialogflow CX agent?

Dialogflow CX builds and hosts the agent, and Cekura tests and monitors it before and after launch.

Cekura is better for:

• Running pre-production simulations of a CX agent across interruptions, background noise, and adversarial callers.
• Monitoring live calls for drop-off points, instruction-following, and compliance checks.

Dialogflow CX is better for:

• Building deterministic flows, Playbooks, and data stores inside Google Cloud.
• Hosting the agent's NLU, speech pipeline, and Google telephony under one bill.

Use both if:

• You run a healthcare scheduling agent under HIPAA, where a missed branch means a patient misses an appointment.
• You change routes or Playbook instructions every week and need a regression suite on each change.

Cekura has a native Gemini Enterprise CX integration for agents built in CX Agent Studio. It connects through a read-only Google service account, places test calls to the agent's phone number, and reads conversation history for monitoring.

Text testing of CX Agent Studio agents is not supported yet. For Flows and Playbooks agents, wrap DetectIntent behind Cekura's custom WebSocket or API chat connection for text, and point Cekura at the agent's phone number or a SIP endpoint for voice.

Production transcripts can flow in through a webhook-based custom integration. That gives you one monitoring view across test and live calls.

Cekura also runs public voice benchmarks at benchmarks.cekura.ai, where we test how leading platforms hold up across repeated runs, not just one clean call.

On the Voice Quality benchmark, Bland topped 9 platforms across 270 calls with a 92.8 score, just ahead of Retell at 92.2. On speech-to-speech, GPT Realtime 2.1 led reliability at 79.3% pass³ over 2,460 calls.

Our Agent Workflow benchmark ran 8 voice agent configurations through the same 82 scenarios three times each. Even the leader, Retell, passed all three runs on 62 of 82 scenarios, so roughly one scenario in four failed at least once.

Dialogflow CX was not in either cohort, but the lesson carries over: one clean test call is not a measurement.

Run the same scenarios against your CX agent on every change, so you see whether it holds up run after run before it takes a live call.

Cekura covers three stages of a CX agent's lifecycle:

• Pre-production: Persona-driven simulations, text-based regression runs on every change, and DTMF testing for keypad paths.
• Infrastructure: Load testing for concurrent calls, with latency and interruption checks.
• Observability: Production call monitoring with LLM-judge and Python metrics, plus PII redaction.

Native integrations work out of the box for Retell, Vapi, ElevenLabs, LiveKit, Pipecat, Bland, and more. Custom connections cover stacks like Dialogflow CX.

Cekura is SOC 2-, HIPAA-, and GDPR-compliant per its trust center. A signed BAA starts on the Startup plan, and audit logs come with Enterprise.

Cekura's pay-as-you-go plan runs $0.25 per voice testing minute, $0.025 per chat reply, and $0.05 per monitored call, with one free seat, then $30 a month per additional seat.

Every plan starts with 300 free credits, about 60 voice minutes. The Startup plan is $500 per month and includes about 2,000 voice testing minutes and 10 seats. Pricing checked October 2026.

Book a demo to put your CX agent through its first simulation suite.

The bottom line on Dialogflow CX pricing

Dialogflow CX pricing is lowest when agents follow structured flows. Flows cost $0.007 per chat request and $0.06 per voice minute, and Playbooks cost 71% more per request and double per second.

Pick the entry point before the plan. I'd start every new CX agent in a Flow. In the example above, a Flow-led agent with generative fallback on 20% of its seconds cost $4,320 a month, and the same traffic started in a Playbook cost $7,200.

Budget the first year against list prices, since the $600 and $1,000 credits run out in days at production volume. Then add requests per turn and test traffic to the model before you sign off on a forecast.

Frequently asked questions

How much does Dialogflow CX cost per month?

Dialogflow CX costs $0.007 per chat request and $0.001 per voice second for Flows agents, with no monthly platform fee. A line handling 20,000 three-minute calls lands around $3,600 a month on Flows and $7,200 on Playbooks.

Is Dialogflow free to use?

No, Dialogflow CX has no ongoing free tier, though new users get $600 in Flows credit and $1,000 in Playbooks credit for 12 months. Dialogflow ES offers a no-charge Trial Edition with a limited audio quota.

How much does a Dialogflow CX voice call cost per minute?

A Dialogflow CX voice call costs $0.06 per minute on Flows and $0.12 per minute on Playbooks. Google bills listening time plus response time per turn, rounds up to the nearest second, and bills interrupted responses in full.

What is the difference between Dialogflow ES and CX pricing?

The main difference between Dialogflow ES and CX pricing is how the meters are split. ES bills $0.002 per text request plus separate audio input, audio output, and phone gateway charges. CX bills $0.007 per text request and bundles speech and telephony into $0.06 per voice minute.

Does generative fallback cost more in Dialogflow CX?

Yes, generative fallback costs more. A turn that reaches it bills at the Playbooks rate of $0.012 per chat request, as one request with no Flows charge stacked on top.

How does CX Agent Studio pricing compare to Dialogflow CX?

CX Agent Studio charges $0.50 per chat or voice session, with evaluations and monitoring included. It costs more than Flows for most conversations. It beats Playbooks only on voice calls of 250 to 300 seconds or chats of about 42 to 50 turns.

Do test conversations count toward the Dialogflow CX bill?

Yes, test conversations sent through the API or a phone integration are runtime requests and bill at standard rates. Google lists design-time requests, like using the visual builder, as free.