# AI agent builder pricing, Copilot Studio, Agentforce, Botpress and CrewAI

Canonical URL: https://industrycontents.com/agent-builders-pricing/
Language: en-GB
Published: 2026-10-14
Structure updated: 2026-10-25

> Dated article snapshot. Canonical web page may receive later updates. Vendor figures retain their stated source-check dates. Illustrations are not product benchmarks.

Published: 2026-10-14 · Structure updated: 2026-10-25. Vendor prices and source checks retain the dates stated below.

## Quick answer

Copilot Studio, Agentforce, Botpress and CrewAI bill different events, so a conversation count alone cannot rank their costs. Price the actual actions, model use, capacity commitments and human work required for the same accepted outcome.

## Key terms

**Copilot Credit**: Copilot Credit is Microsoft Copilot Studio’s usage unit, consumed at different rates by agent features.

**Flex Credit**: Flex Credit is Salesforce Agentforce’s consumption unit; different action types have specified rates.

**Accepted outcome**: Accepted outcome is a result that passes the buyer’s stated quality and completion checks.

## AI agent builder pricing, plans and billing units

Copilot Studio meters Copilot Credits, Agentforce offers several meters, Botpress counts conversations and CrewAI prices workflow capacity.

Each product bills in a different unit, so the table below puts them side by side.

| Comparison point | Copilot Studio | Agentforce | Botpress | CrewAI |
| --- | --- | --- | --- | --- |
| Billing unit | Credits | Credits, or $2 a conversation | Conversations | Workflow executions |
| Free start | Trial | Trial | Free plan, 100 conversations | Free, 50 executions a month |
| Entry paid price | $200 per month for 25,000 credits, billed annually ($0.008 capacity value) | $500 per 100,000 Flex Credits ($0.005 each) | Plus $150 a month billed annually, 250 conversations | Custom |
| Cost of one action or answer | Generative answer 2 credits, agent action 5 credits (pack value $0.04; pay-as-you-go $0.05) | Standard action 20 credits ($0.10) | Included in the conversation, with $0.10 of AI usage per conversation covered | Sized to workflow |
| Extra usage | More credits, pay as you go | Contracted rate, billed in arrears | $65 per 100 conversations (Plus), $50 per 100 (Team) | Flexible overage in contract |
| Team and seats | Employee use included for licensed Microsoft 365 Copilot users | $125 per user add on, unmetered | 3 seats on Plus, unlimited on Team ($750 a month) | Enterprise SSO and RBAC |
| At the limit | Prepaid capacity: custom-agent enforcement at 125%; new agent-flow runs blocked when capacity is exhausted. Pay-as-you-go differs. | No overage penalty, billed after | Packs of 100 added automatically at 95% of quota | Not stated |

Caption: Start with how each platform charges. The table sets out Copilot Studio, Agentforce, Botpress, CrewAI.

## AI agent builder costs at 250, 1,500 and 6,000 conversations

The conversation-budget table assumes one generated answer and one agent action; it is not a general cost per support conversation.

For a deliberately simple counting scenario, we assumed one generated answer and one billable agent action per conversation. These are billing assumptions, not evidence that the products implement identical work. On Copilot Studio that comes to 7 credits, $0.056 of pack capacity or $0.07 on pay as you go. On Agentforce Flex it means two standard actions at 20 credits each, or $0.20. The figures below are estimates built from published prices. They leave out extra licences, grounding, retries and setup, so use them to shortlist options and then test your workflow.

| Conversations a month | Copilot Studio | Agentforce (Flex) | Agentforce ($2 each) | Botpress |
| --- | --- | --- | --- | --- |
| 250 | $17.50 pay as you go; $200 for one pack | $50 | $500 | $150 (Plus) |
| 1,500 | $105 pay as you go; $200 for one pack | $300 | $3,000 | $750 (Team) |
| 6,000 | $420 pay as you go; $400 for two packs | $1,200 | $12,000 | $3,000 nominal capacity (Team plus 45 packs); recharge timing and AI spend extra |

Caption: Put the same support workload through the numbers. The table sets out Conversations a month, Copilot Studio, Agentforce (Flex), Agentforce ($2 each), Botpress.

Copilot Studio has two different purchase models. At seven credits per conversation, pay as you go at $0.01 per credit costs $17.50, $105 and $420 for these volumes. A $200 monthly pack contains 25,000 credits, so pack purchases cost $200, $200 and $400 respectively, with unused capacity in the first two examples. The $0.008 pack unit value is not the pay-as-you-go rate. Microsoft’s [licensing guidance](https://www.microsoft.com/licensing/guidance/Microsoft-Copilot-Studio), rechecked 25 October 2026, lists the pack at $200 per month billed annually, with no monthly rollover. The pack examples are monthly equivalents, not a commitment-free purchase.

## One conversation, several bills

Volume is only half the forecast. A knowledge answer, a grounded account lookup and a multi-step case update can all be called a conversation while consuming different resources. The seven-credit example above deliberately leaves out that variation.

| Assumed flow | Credits per conversation | Credits at 6,000 conversations | Pay-as-you-go at $0.01 |
| --- | --- | --- | --- |
| One generative answer + one agent action | 2 + 5 = 7 | 42,000 | $420 |
| Two answers + two actions + one tenant graph grounding event | 4 + 10 + 10 = 24 | 144,000 | $1,440 |
| The 24-credit flow repeated once in full | 48 | 288,000 | $2,880 |

Caption: IC arithmetic from Microsoft’s listed feature rates. These are invented flow compositions, not measured averages. Excludes other AI tools, voice, external models and additional licensing.

The point is the sensitivity, not a prediction that every retry doubles a bill. A recovery path might repeat one action rather than the full flow. Capture the actual events in a pilot and average their cost across successful and unsuccessful attempts.

### Pack or meter?

A $200 pack buys 25,000 credits a month, while pay-as-you-go costs $0.01 each. In isolation, one pack becomes cheaper than pure metering above 20,000 credits and covers up to 25,000. Above that, compare another pack with metered excess rather than assuming every extra credit requires a full pack. Factor in the annual pack commitment and quiet months: unused monthly capacity does not carry forward.

## Agentforce action depth

At the listed 20 Flex Credits per standard action and $500 per 100,000 credits, one action has a $0.10 list-rate value. Two actions use $0.20; ten use $1; twenty use $2. The $2 conversation option reaches arithmetic parity at twenty standard actions, but that is not a purchasing recommendation. Eligibility, contract terms, voice and other services can change the comparison, and Salesforce says Flex and Conversation pricing cannot coexist in the same organisation.

The existing $50 and $300 Flex examples describe consumed usage at list-rate value. They do not establish the minimum purchase or invoice for your buying model. Get the applicable pre-purchase, commitment or pay-as-you-go terms in writing.

## Botpress has two meters

Conversation capacity and AI spend need separate lines. The Plus and Team plans include AI usage, but heavier model use can exceed that allowance. Botpress’s FAQ says paid workspaces automatically buy a $10 AI credit grant at 95% of the AI-spend limit; unused grant value can carry for up to 12 months from purchase month-end. That differs from conversation packs, whose unused conversations expire monthly.

Budget for peak months and check retained packs after traffic falls. A temporary spike should not become a permanent line item just because nobody visits billing settings.

## The delivery bill

CrewAI’s hosted free allowance is not a price for running an entire production service. Ask the Enterprise quote to separate platform capacity, model and tool charges, infrastructure and engineering support. Compare the responsibilities left with your team rather than treating custom pricing as either cheap or expensive.

For any builder, add implementation and operations to consumption. With invented inputs, a $300 monthly platform bill plus four maintenance hours at $75 is $600 before model extras. A $2,400 setup project allocated across 12 months adds another $200. If 800 outcomes pass review, that is $1 per accepted outcome. None of those inputs is a vendor quote.

## Check your existing software before comparing rates

If your team manages its data and permissions in Microsoft or Salesforce, begin there. You may save a good deal of integration work even when the usage rate is higher. Botpress gives a smaller team the most direct route to a webchat or messaging agent, and CrewAI makes sense when your developers want control of the workflow itself. Settle who will maintain the agent after launch too, because that person's time belongs in the cost comparison.

## How the bill grows when your agent gets busy

Botpress counts any exchange with at least two end user messages as a conversation, counts anonymous sessions separately, and counts three minutes of voice as one. Unused Botpress pack conversations expire monthly, but purchased packs persist into following months until removed in billing settings. Conversation-pack auto-recharge cannot be turned off. Copilot Studio’s 125% custom-agent threshold applies to prepaid capacity; agent flows and pay-as-you-go have different enforcement rules. Agentforce's Flex and Conversations options cannot be mixed. CrewAI's free tier stops at 50 executions, which suits testing and will be outgrown quickly by a live support queue.

## Budget for the whole conversation

At 6,000 conversations, this simplified workload needs 42,000 Copilot Credits, $420 on pay as you go or $400 for two monthly packs. The Botpress Team example is $3,000 of nominal conversation capacity; it is not a guaranteed invoice because automatic purchases occur before capacity is fully consumed and AI spending can add charges. These totals buy different implementations and leave out extra licences, AI spending above included amounts, grounding and integration work. Ask each vendor to price the same representative flow, including failed attempts and handoffs.

## Two more builders that require a quote

Relevance AI and Voiceflow both build business agents, and neither shows a plan price on its pricing page. Relevance AI lists Enterprise only. Voiceflow splits into agency and business tiers and routes pricing through a demo. Expect to request a quote from both.

## How to benchmark agent builders on the same support flow

- Pick one real support flow and write down its steps.

- Use the same sample and number of conversations for every contender. If a free allowance permits only 50, run 50 across all tools or buy enough capacity for the full sample. Extrapolate only after stating the sample size and assumptions.

- Include normal questions, missing records, denied permissions, ambiguous requests, API timeouts and human handoffs. Record each billed event and the recovery path.

- Score the result independently: correct answer, correct record update, safe refusal or appropriate handoff. A completed run is not automatically an accepted outcome.

- Scale by your expected mix of simple and complex work, then add licences, model or tool charges, setup and maintenance. Check both normal and peak months.

## Get four answers in writing before signing

- What does our flow cost per conversation, all in?

- What do we already have that lowers the price?

- What happens at the limit, and who is told?

- What counts as a conversation, an action or an execution?

## Questions to settle before you choose

### Is Copilot Studio the cheapest AI agent builder for a small business?

There is no universal cheapest builder. Our seven-credit Copilot Studio scenario costs $17.50 on pay as you go at 250 conversations; choosing capacity packs instead requires a $200 purchase. Botpress Plus is $150 on annual billing for 250 conversations. Feature requirements, licences and implementation can reverse the comparison.

### Does Botpress charge per message?

No. It charges per conversation, meaning any exchange with at least two end user messages in the billing month.

### Is CrewAI free?

CrewAI has a free plan with 50 workflow executions a month. Larger use moves to Enterprise with custom pricing.

### What should an agent-builder pilot include?

Run the same support flow on each shortlisted builder. Record platform units, model usage, tool calls, failed attempts and human review time. Compare accepted outcomes and the full deployment bill.

[Check the build-or-buy decision](/guides/build-or-buy-ai-agent/). Use the dated sources in this article for price context; the comparison method is an editorial framework, not a hands-on product benchmark.

Depth update, 25 October 2026: rechecked [Microsoft feature rates and enforcement](https://learn.microsoft.com/en-us/microsoft-copilot-studio/requirements-messages-management), [licensing terms](https://www.microsoft.com/licensing/guidance/Microsoft-Copilot-Studio), [Salesforce pricing](https://www.salesforce.com/agentforce/pricing/), [Botpress pricing and recharge FAQ](https://botpress.com/pricing) and [CrewAI plans](https://crewai.com/pricing). New flow, retry and labour examples are editorial assumptions, not hands-on benchmarks. Relevance AI and Voiceflow retain their earlier check dates.

**Sources**

- Microsoft Learn, [Billing rates and management](https://learn.microsoft.com/en-us/microsoft-copilot-studio/requirements-messages-management), read 12 October 2026.

- Salesforce, [Agentforce pricing](https://www.salesforce.com/agentforce/pricing/), read 12 October 2026.

- Botpress, [pricing](https://botpress.com/pricing), checked 14 October 2026.

- CrewAI, [pricing](https://crewai.com/pricing), checked 14 October 2026.

- Relevance AI, [pricing](https://relevanceai.com/pricing), and Voiceflow, [pricing](https://www.voiceflow.com/pricing), checked 14 October 2026.

Worked examples are IC arithmetic from published rates and stated assumptions. Read [Copilot Studio pricing](/guides/copilot-studio-pricing-licensing/) and [Agentforce pricing](/guides/agentforce-pricing-flex-credits/), see the [AI Agent Pricing Tracker](/ic-lab/ai-agent-pricing-tracker/), or go back to [all comparisons](/ic-lab/compare/). [Send a correction](/ic-lab/ai-agent-pricing-tracker/#add-your-company).

IC Lab · AI Agent Pricing Tracker

Take the pricing rows into your budget.

Compare source-linked rows in the tracker. Open the tracker and find the CSV export. The tracker offers a CSV export with a TalkHead signup. TalkHead is our weekly newsletter on AI in business.

**Related** Copilot Studio pricing · Agentforce pricing · [Copilot Studio governance](/guides/copilot-studio-governance-controls/) · [Build or buy an AI agent](/guides/build-or-buy-ai-agent/) · AI Agent Pricing Tracker

## Keep exploring the evidence

**[The billing unit changes cost](https://industrycontents.com/ai-agents-breaking-saas-user-definition/)** An AI worker can consume credits without being a licensed employee. Our analysis examines how that changes enterprise software economics.

For the infrastructure behind increasingly capable AI agents, read our [analysis of Cloudflare MCP Code Mode and the future of agent tool calling](https://industrycontents.com/cloudflare-code-mode-mcp-explained/).

Our [UiPath and Warmwind analysis](https://industrycontents.com/uipath-and-warmwind-agentic-automation/) examines who builds and supervises these workflows. Platform pricing is only one part of that deployment decision.