respond.io AI Agents are included from the Growth plan upwards, at $159 a month billed yearly ($199 billed monthly). Starter does not include them. Extra AI credits cost $15 per 1,000, switch on automatically, and stop at 200% of your plan's allowance. A Growth plan with 1,000 contacts includes 10,000 credits each month.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowThat headline hides the part that decides your bill: each AI Agent reply uses 2 to 5 credits, so 10,000 credits stretch to somewhere between 2,000 and 5,000 replies. A business with chatty customers can use its allowance before the month ends. The sections below show how to estimate your own use, test cheaply, and pick between Growth and Advanced.
Which respond.io plan includes AI Agents, and how many credits does each get?
respond.io includes AI Agents from the Growth plan upwards; Starter does not. Respond.io's pricing page lists four plans. The figures below were read from the page in November 2026; the page itself warns that some allowances start at a minimum and grow with your contact tier.
| Plan | Billed yearly (per month) | Billed monthly | AI Agents | Included credits |
|---|---|---|---|---|
| Starter | $79 | $99 | No (AI Prompt and AI Assist only) | 5,000, fixed |
| Growth | $159 | $199 | Yes | 10 per contact, so 10,000 at 1,000 contacts |
| Advanced | $279 | $349 | Yes, plus AI voice agents | 20 per contact, so 20,000 at 1,000 contacts |
| Enterprise | Custom | Custom | Yes | 40 per contact, so 40,000 at 1,000 contacts |
The help-centre article on credits states the multipliers: Growth includes 10 credits per monthly active contact, Advanced 20 and Enterprise 40, scaling with your contact tier up to 10,000 contacts. Starter gets a fixed amount that does not scale. The pricing page also says AI usage is included at no extra cost on the self-service Growth, Advanced and Enterprise plans, subject to a fair use policy, and that credits refresh at the start of each billing cycle.
One definition matters before any sum: a monthly active contact is someone you have talked to over calls or chat during the billing month. Contacts who message you but never get a reply from your team or agent are a grey area, so confirm how your account counts them in the Billing and Usage screen after the trial.
How many credits does each respond.io AI action use?
Each respond.io AI action costs a set number of credits, and the credits help page gives a per-feature list to build your estimate on.
- AI Agent reply: 2 to 5 credits per response.
- AI Prompt, AI Translate, AI Assist, AI Chat Summary: 1 credit each.
- Closing notes and sentiment: 1 credit each, so 2 per closed conversation if you switch both on.
- Call and audio transcription: about 2 credits per minute.
- Voice AI Agent: about 70 credits per minute.
Conversations marked Unavailable use no credits. That is useful for spam: mark the obvious junk and it stops costing you.
Notice how lopsided voice is. Ten minutes of a voice agent costs about 700 credits, which is more than the entire written-reply budget of many small accounts for a day. Treat voice as its own line in the budget, and if the voice line looks too expensive on credits, compare a dedicated phone product before enabling it. Our tutorial on trialling an AI answering service shows how to judge those.
A pet shop with 1,000 contacts, worked through
An illustration, with made-up but plausible figures. A pet shop sells food, grooming slots and small-animal supplies, and most of its customers write on WhatsApp and Instagram. It has about 1,000 active contacts in a month and wants the AI Agent to answer opening hours, delivery questions and stock checks, handing anything awkward to staff.
Step one is to estimate replies per contact. Say 800 of those contacts get an AI answer during the month, with an average of five AI Agent replies each. That is 4,000 replies. At the low end of 2 credits each, the use is 8,000 credits; at the high end of 5 credits, it is 20,000. The answer depends on reply length and how often the agent searches its knowledge, which only your own test will show.
| Scenario | Credits used | Growth (10,000 included) | Cost above $159 |
|---|---|---|---|
| Light: 2 credits a reply | 8,000 | Inside allowance | $0 |
| Typical: 3.5 credits a reply | 14,000 | 4,000 over, 4 blocks | $60 |
| Heavy: 5 credits a reply | 20,000 | 10,000 over, 10 blocks, exactly at the 200% cap | $150 |
Compare the heavy case with Advanced. Growth at 20,000 credits costs $159 plus $150, which is $309 a month. Advanced is $279 and already includes 20,000 credits, so it is $30 cheaper and gives headroom under the cap. The crossover is easy to find: Advanced costs $120 more than Growth on yearly billing, and $120 buys eight 1,000-credit blocks, so once you expect to buy more than eight blocks a month, Advanced is cheaper. In the typical case the shop buys four blocks ($60), so Growth plus overage wins.
Two extra costs sit outside all of this. Meta, the company behind WhatsApp, charges its own conversation fees which are not part of any respond.io subscription; respond.io's page says WhatsApp fees are billed by Meta on usage and you can top up WhatsApp credit on the platform. Extra users are charged per seat: $12 a month on Starter, $20 on Growth and $24 on Advanced. Our WhatsApp pricing tutorial goes through those fees.
What happens when respond.io AI credits run out?
On current paid plans, when respond.io AI credits run out, on-demand credits switch on automatically at $15 per 1,000 until total use reaches 200% of your allowance. The rules, as stated on respond.io's FAQ and help pages:
- When your included credits run out, on-demand credits switch on automatically on current paid plans, so the agent keeps replying.
- They cost $15 per 1,000, rounded up to full blocks. The FAQ's own example is 1,500 extra credits costing $30.
- The charge lands on your subscription invoice when the billing cycle resets.
- Total use, included plus on-demand, is capped at 200% of your allowance. After the cap, the help page says AI features pause until the next cycle.
- On-demand is not available on Trial or Legacy plans.
- Email and in-app alerts warn you as use climbs, and the Billing and Usage page shows a progress bar and daily chart.
Trial and legacy plans (Team, Business, Legacy Enterprise) do not offer on-demand credits: on those, AI features pause when the included credits run out. Watch your own Billing and Usage page in the first month to see which behaviour your account follows.
Rounding up to full blocks means a single extra credit costs $15 in that month. It is a small point, but it explains an invoice that looks $15 higher than a straight per-credit sum.
Can you test respond.io AI Agents before you upgrade?
You can test respond.io AI Agents through the 7-day trial, because Starter customers cannot test them inside their own plan. The pricing page says the trial gives Growth features for 5 users and 1,000 monthly active contacts, with broadcasts excluded. Use those seven days as a measurement, not a demo.
A workable plan:
- Day 1: connect one channel and load your top 20 questions into the agent's knowledge. Our tutorial on training the agent on your FAQs covers this.
- Days 2 and 3: have two staff send 30 real past customer messages each. Note the credits used per conversation from the usage chart.
- Days 4 to 6: send the agent difficult cases: angry messages, a question with no answer in the knowledge, and a request to speak to a person. Check each handover.
- Day 7: divide credits used by AI replies sent. That ratio is your own credits-per-reply figure, the number the table above guessed at.
Record your results in a small sheet like this one. The numbers are illustrative.
Test messages sent: 60
AI Agent replies: 214
Credits used (usage chart): 612
Credits per reply: 2.9
Replies handed to a human: 9 of 60 conversations
Wrong or risky answers: 3 (all about refunds)
Projected monthly replies: 4,000
Projected credits: 4,000 x 2.9 = 11,600
Plan check: Growth 10,000 + 2 blocks ($30)
The "wrong or risky" row matters as much as the credit row. Three refund answers out of sixty is a reason to fix the knowledge base or add a rule sending refund questions to a person, not a reason to upgrade the plan.
If you want help working out which plan fits how your business actually talks to customers, I offer one-to-one AI consultations for small businesses.
A salon, a gym and a cafe at different message volumes
The right plan changes with how people use the channel. Three short illustrations, with assumed figures:
A hair salon with 300 contacts. Mostly booking requests, around two agent replies per contact. That is about 600 replies, perhaps 1,800 credits. Growth's entry tier starts at 1,000 contacts with 10,000 credits, so overage never comes into it. The question is whether the agent is worth $159 at all when a booking tool and an auto-reply would do; see our guide to one AI inbox for WhatsApp, Instagram and Facebook for lighter options.
A gym with 2,000 contacts. Heavy on membership questions, class times and freezes. If each contact generates three AI replies at 3 credits, that is 18,000 credits. The allowance is tied to your contact tier, so check what the 2,000-contact tier includes on the pricing page rather than assuming it doubles. Contact overage on Growth is billed at $12 per 100 contacts, and on Advanced at $15 per 100, so a 1,000-contact overshoot costs $120 on Growth.
A cafe with 500 contacts and a voice line. The owner wants the AI to answer phone enquiries about tables. At 70 credits per minute, a 2-minute call is 140 credits. Forty calls a month is 5,600 credits before a single written message. That would need Advanced, which is where AI voice agents appear, and it may be cheaper to use a dedicated phone answering product.
Where AI credit budgets leak
Some patterns inflate credit use without improving customer service. Watch for them in the daily usage chart:
- Greeting loops. The agent replies to every "ok" and "thanks", spending 2 credits on a message that needed none. Add a rule to stay silent after a closing word.
- Automated senders. Delivery notifications and spam contacts that trigger replies. Mark them Unavailable so they use no credits.
- Extras switched on by default. Closing notes plus sentiment add 2 credits to every closed conversation. On 1,000 conversations that is 2,000 credits, a fifth of the Growth allowance, for data nobody reads.
- Long knowledge answers. An agent that pulls a long article and rewrites it can sit at the 5-credit end. Tighten the source text.
- Transcribing calls you never review. About 2 credits a minute is small per call and large across a busy line.
Cut those first and the plan you needed often drops a tier.
How do you cut respond.io AI credit use without hurting service?
The cheapest reply is the one the AI Agent does not need to write, because credits are spent per reply. Four rules, with an illustrative before and after for a pet shop's WhatsApp line.
Rule 1: answer fixed questions with fixed text. Opening hours, the returns address and the delivery cut-off do not need a model. A workflow step or a saved reply costs nothing in AI credits, while the agent costs 2 to 5 each time it answers the same question. If 30% of your AI replies are the same five questions, moving them out saves roughly a third of the budget.
Rule 2: stay silent after a closing word. Compare two endings to the same chat. Before: customer says "Thanks!", the agent replies "You're very welcome! Is there anything else I can help with today?" (one reply, perhaps 2 credits). After: the rule treats "thanks", "ok" and a thumbs-up as the end of the conversation and sends nothing. Across 1,000 conversations that is up to 2,000 credits kept, which is a fifth of the Growth allowance.
Rule 3: hand over early. A conversation that loops through six AI replies before reaching a person costs 12 to 30 credits and annoys the customer. Set the agent to hand over after two failed attempts, or at once for words like "refund", "complaint" or "manager". The customer gets a human sooner and you pay for fewer replies.
Rule 4: keep prompts short. The agent's instructions are read on every reply. A three-page rulebook of exceptions is harder to follow than a one-page one, and long source articles in the knowledge base push replies towards the 5-credit end. A starting instruction like the one below is enough for a first week; the wording is illustrative.
You answer WhatsApp messages for a small pet shop.
Use only the knowledge base. If the answer is not there, say you will ask a colleague and hand over.
Keep replies under 60 words. No emojis.
Hand over at once for: refunds, complaints, injured or unwell animals, anything about an order over $200.
Do not reply to "thanks", "ok" or a single emoji.
That last line is a credit rule as much as a style rule, and the "injured or unwell animals" line is a safety rule: a shop should never let an AI improvise advice about a sick pet. Read a sample of real conversations each week for the first month and adjust the list.
How do you review respond.io AI credit use each month?
Review AI credit use once a month, on the first working day, from the Billing and Usage page. Once the agent is live, set a diary reminder, then copy five numbers into a sheet: included credits, credits used, on-demand credits bought, the on-demand charge, and active contacts. Then work out two ratios: credits used divided by active contacts, and on-demand spend divided by the plan fee. The first tells you whether your use per customer is rising; the second tells you whether you are drifting towards an upgrade.
An illustration of a three-month log for the same pet shop:
Month Contacts Credits used Included Extra bought Extra cost
Jan 980 8,900 10,000 0 $0
Feb 1,040 11,800 10,000 2 blocks $30
Mar 1,120 14,600 10,000 5 blocks $75
The trend is the signal, not any single month. Extra spend of $0, $30 and $75 against a $159 plan fee is rising fast; if April lands near $120, Advanced at $279 (an extra $120 over Growth) becomes the cheaper option, because $120 buys eight blocks of 1,000 credits. Contacts also crossed 1,000 in February, so check whether a higher contact tier changed the allowance, since on Growth it is 10 credits per contact. A shop that notices this in March makes a calm decision; one that notices on an invoice makes a rushed one.
Also open a handful of conversations each month where the agent handed over to a person and ask whether it should have answered, and a handful where it answered and should have handed over. Those two lists are the best source of new rules for the instruction above, and they cost nothing in credits.
If that sheet is where you track credits, The Complete Excel Course with AI Integration on Coding Liquids may be useful.
The upgrade rule I would use
Start on Growth if you want AI Agents at all, because Starter cannot run them. Run the 7-day measurement and compute credits per reply. Then apply these three rules:
- If projected credits are within the included allowance, stay on Growth and set an alert at 80% usage.
- If you expect to buy up to about eight blocks of on-demand credits a month, stay on Growth and accept the overage, since it is cheaper than Advanced's extra $120.
- If you expect to buy more than that, or you need voice agents, move to Advanced and recheck the numbers after a month.
Contact tiers and credit allowances change, so use the figures here as a method and confirm the live numbers on respond.io's pricing page before paying. For the full plan-by-plan cost at 1,000 contacts, read respond.io pricing for 1,000 contacts, and if you are still comparing tools, WATI vs respond.io for a three-person team lays out the alternative.
Credit and plan questions people ask next
Does respond.io Starter include AI Agents?
No. Respond.io's pricing page marks AI Agents as not included on Starter, which offers AI Prompt and AI Assist only. AI Agents begin on Growth, and Advanced and Enterprise include them too. Starter does carry a fixed 5,000-credit allowance for the lighter AI tools, so you can still draft replies with AI.
What happens when the AI credits run out?
On current paid plans, on-demand credits switch on automatically at $15 per 1,000, rounded up to full blocks and billed when the cycle resets. Total use is capped at 200% of your allowance; after that, respond.io's help page says AI features pause until the next cycle.
Can I try AI Agents before paying?
Yes, through a 7-day free trial. Respond.io advertises a 7-day free trial with Growth features for 5 users and 1,000 monthly active contacts, with broadcasts excluded. Its help page says on-demand credits are not available on Trial plans, so a trial shows behaviour but not overage billing.
Do voice calls use the same credits?
Yes, but at a much higher rate. Respond.io's help page lists about 70 credits per minute for a Voice AI Agent and about 2 credits per minute for call transcription, against 2 to 5 credits for a typical written AI Agent reply.
Further reads
Browse all AI-for-business tutorials
Sources: respond.io pricing page, respond.io help-centre AI Credits article and AI Agents credits FAQ, and respond.io's pricing explainer, read in November 2026.
