No-code AI agent builders: which one should a non-developer use?
Seven no-code agent builders compared on the billing unit, the cost of one five-step agent at 100, 1,000 and 10,000 runs a month (tiers read 2026-11-09), human-approval controls and where your data is processed.
Independent editorial content. Links to software go straight to the vendors. We are not in any affiliate programme yet and earn nothing from these links; if that changes, this note will say so. Rankings are never sold. See our methodology. Methodology
| Tool | Billing unit | Entry paid price (checked 2026-11-09) | Free plan | Approval and guardrails | Who should not buy | |
|---|---|---|---|---|---|---|
| Zapier Agents Agents inside 9,000+ app automations; low ceiling | Activity (10 per run Free, 40 paid) | Agents Pro ~USD 33.33/mo annual for 1,500 activities | 400 activities/mo | Per-run cap; asks before continuing; knowledge from docs and links | Agents that run more than ~300 five-step runs a month | Visit |
| Make AI Agents Visual scenarios; prices not readable today | Credit (1 per module action) | see pricing (plan cards did not render) | 1,000 credits/mo | Manual approvals, stop points, reasoning panel | Teams that need a quotable price without logging in | Visit |
| n8n Per-run billing, approval for tool calls, Frankfurt | Workflow execution | Starter EUR 20/mo annual (2,500 executions); Pro EUR 50 (10,000) | 14-day trial (1,000 executions); self-hosted Community Edition | Human approval for tool calls; AI usage controls; evaluations | People who want dropdowns, not a node canvas | Visit |
| Lindy AI employee for inbox, calendar and calls | Credit (2–2,500 per task) | Team USD 29.99/seat/mo, 3,000 pooled credits; top-up USD 10/1,000 | USD 50 of credits, expire after 7 days | Approvals built in for outside-impact actions; editable memory | Users with many deep-work tasks on a fixed budget | |
| Dify Cloud Knowledge-base chat agents; self-host later | Message credit (by model type) | Professional USD 590/workspace/year (5,000 credits/mo) | Sandbox: 200 credits, 5 apps, 50 documents | Knowledge base quotas; approval not stated | Teams needing a stated EU region | |
| Voiceflow Agent builder sold through sales | Usage-based with auto top-ups | Request pricing | Agency trial, no card | Global instructions and guardrails | Individuals wanting a public price | |
| Relevance AI Enterprise agent workforce | Custom Vendor Credits | Talk to sales | None shown | Agent evaluations, SSO, RBAC, audit logs | Small teams |
Prices and features checked: 2026-11-09. Prices in the currency the vendor shows, checked on the date shown; we do not convert.
Verdict by user type
Every tool on this page lets someone who does not write code describe an agent, connect it to apps and let it act. What differs is the unit on the bill, the ceiling on the free plan, and what happens when the agent is wrong. We organise the comparison around those three things, because the feature lists have converged and the bills have not.
Our criteria, stated first: the billing unit and what one unit covers; the cost of one reference agent at 100, 1,000 and 10,000 runs a month, computed from the vendor's published tiers on 2026-11-09; the controls for memory, knowledge and human approval as documented; and any statement about where data is processed. Zapier, Make and n8n have affiliate programmes and are in our catalogue; Lindy, Dify, Voiceflow and Relevance AI are not. The order does not depend on commission.
We have not built the same agent in every tool and timed it. Our plan called for a three-tool build log; this page compares documented pricing and features and marks the cost model's assumptions so you can rerun it with your own numbers.
| You are… | Pick | Why |
|---|---|---|
| A non-developer whose data already lives in SaaS apps | Zapier Agents | 9,000+ apps, 400 free activities a month, hard cap per run, the agent asks before continuing |
| A team that needs a stated EU region and a human approval step | n8n Cloud | Per-run billing, "human approval for tool calls", data stored in Frankfurt |
| A manager who wants one assistant for inbox, calendar and calls | Lindy | Credit-based, approvals built in, 1,500+ integrations and MCP |
| A team building a knowledge-base chat agent on its own documents | Dify Cloud | Document and storage quotas per plan, free Sandbox, self-host later under its licence |
| A visual thinker who already uses Make | Make AI Agents | "Available on all plans", manual approvals and stop points; prices did not render for us |
| A customer-service team buying through procurement | Voiceflow or Relevance AI | Both sell to businesses through sales conversations; neither published self-serve prices we could read |
What "no-code agent" means in each tool
The word covers four architectures, and knowing which one you are buying predicts the bill.
- Agent as a step inside an automation (Zapier Agents, Make AI Agents, n8n). The agent is a node in a workflow you already understand; the rest of the workflow is deterministic. Zapier bills the agent in "activities, not tasks": a trigger, a knowledge-source answer, an action, a web browse or scrape, a web search and a Chrome-extension message each count as one [1][2]. Make bills everything in credits, "most actions cost 1 credit", with the agent "available on all plans" [3][4]. n8n bills the whole run as one execution "regardless of complexity" [5].
- Agent as an employee (Lindy). You describe a role; the tool handles inbox, calendar, calls and integrations and asks for approval on "anything with outside impact". Credits measure the work: "Everyday Asks 2–250 credits, Deep work 250–1,000, Big Builds 1,000–2,500" [6].
- Agent as an app (Dify, Voiceflow). You build a chat or voice application on a canvas, attach a knowledge base, and expose it to users. Dify bills message credits that depend on the model used; Voiceflow describes "usage-based billing with optional auto top-ups" [7][8].
- Agent as a workforce (Relevance AI). Multiple agents with tools, evaluations, SSO, RBAC and audit logs, sold as an enterprise bundle with "Custom Vendor Credits" [9].
Demand for this category is new and steep: "no code ai agents builder" is searched about 5,400 times a month in the US with a keyword difficulty of 10, "ai agents builder" 6,600, "build ai agents" 1,900 (DataForSEO, retrieved 2026-11-09). In our Future Demand Radar the parent category, "AI agents and delegated work", scores 59.7 out of 100 in the baseline scenario (band 55.2–64.1) and is classed "build now"; our baseline expects the builder segment to follow the parent. That is a forecast with stated uncertainty, not a fact.
Comparison: unit, free plan, controls, residency
| Tool | Billing unit | Free plan (checked 2026-11-09) | Entry paid price | Memory, knowledge, approval | Data residency stated |
|---|---|---|---|---|---|
| Zapier Agents | Activity (10 per run Free, 40 paid) | 400 activities/mo | Agents Pro ~USD 33.33/mo annual, 1,500 activities | Knowledge from "FAQs, docs and public links"; asks before continuing at the cap; memory not stated | Not stated |
| Make AI Agents | Credit (1 per module action) | 1,000 credits/mo | see pricing (plan cards did not render) | "Set clear rules, add manual approvals, or stop the Agent at specific points"; reasoning panel | Not stated |
| n8n | Workflow execution | 14-day Cloud trial, 1,000 executions; Community Edition self-hosted | Starter EUR 20/mo annual, 2,500 executions, 1,600 Assistant credits | "Human approval for tool calls"; AI usage controls; evaluations | EU, Frankfurt, for hosted plans |
| Lindy | Credit (2–2,500 per task) | USD 50 of credits, expire after 7 days | Team USD 29.99/seat/mo, 3,000 pooled credits; top-up USD 10 per 1,000 | "Remembers you. Yours to edit"; "Approvals built in" | SOC 2 and GDPR compliance stated; region not stated |
| Dify Cloud | Message credit (by model type) | Sandbox: 200 credits, 5 apps, 50 documents | Professional USD 590/workspace/year (about USD 49/mo), 5,000 credits/mo | Knowledge base with document quotas; approval not stated | Not stated; GDPR, SOC 2, ISO 27001 badges |
| Voiceflow | Usage-based, auto top-ups | Free trial for agencies, no card | Request pricing | "Global instructions and guardrails"; approval not stated | Not stated |
| Relevance AI | Custom Vendor Credits | None shown | Talk to sales | Agent evaluations, SSO, RBAC, audit logs | Not stated |
Sources: [1][2][3][4][5][6][7][8][9].
Cost at 100, 1,000 and 10,000 runs a month
The reference agent: a trigger (a form or inbox event), one knowledge-base lookup, one web search, two actions in connected apps. Five steps. We state the unit cost per run for each tool, then the lowest published tier that covers the month, annual billing, as read on 2026-11-09. Where the tiers run out we say so rather than extrapolate.
| Runs/month | Zapier Agents (5 activities per run) | n8n (1 execution per run) | Lindy (assume 50 credits per run, an "everyday ask") | Dify Cloud (assume 1 message credit per run) | Make, Voiceflow, Relevance AI |
|---|---|---|---|---|---|
| 100 | 500 activities → above Free (400) → Agents Pro ~USD 33.33 | 100 executions → Starter EUR 20 (2,500 included) | 5,000 credits → Team 3,000 plus 2 top-ups → about USD 50 per seat | 100 credits → Sandbox, USD 0 | see pricing |
| 1,000 | 5,000 activities → above Agents Pro (1,500); Enterprise is "coming soon" with custom pricing → no published price | 1,000 executions → Starter EUR 20 | 50,000 credits → a higher Lindy tier; the page lists tiers up to 1.75 million credits without a per-credit rate we can quote → see pricing | 1,000 credits → Professional USD 590/year (about USD 49/mo, 5,000 credits) | see pricing |
| 10,000 | exceeds all published Agents tiers | 10,000 executions → Pro EUR 50 (10,000 included) | see pricing | 10,000 credits → Team USD 1,590/year (about USD 132/mo, 10,000 credits), or switch to your own API key | see pricing |
Three cautions. First, the Zapier column is a ceiling problem, not a price problem: at 1,500 activities a month the paid tier covers 300 of our five-step runs, so a busy agent outgrows it fast, and the page offered only "coming soon" above that [1]. Second, the Lindy assumption is generous; a "deep work" task at 250–1,000 credits would multiply the bill by five to twenty, and billing is per active seat [6]. Third, Dify's message credits are "consumed based on the model type", so one run may cost more than one credit; the escape hatch is that "once they're used up, you can switch to your own API key" [7], which turns Dify into a per-seat platform fee plus your provider bill.
n8n's column is flat because an execution is one run however many steps it has, including the model calls inside the agent node [5]; the model provider bills those separately unless you use the included Assistant credits. That is also why n8n is cheapest at 10,000 runs and merely ordinary at 100, where the EUR 20 floor applies.
Reliability and error handling
What a tool does when the agent is wrong matters more than what it does when it is right.
- Zapier Agents stops at 10 activities per run on Free and 40 on paid plans and "asks before continuing" [2]. That cap is a budget guard and a loop guard at once.
- Make documents "manual approvals" and stop points and shows a reasoning panel for each step [4]. Routers and error-handler modules are not counted as credits [3].
- n8n ships "human approval for tool calls" on Starter, Pro and Enterprise, plus light evaluations on all plans and metric-based evaluations from Pro [5]. Saved-execution history is the debugging tool; Starter keeps fewer runs than Pro.
- Lindy waits for approval on "anything with outside impact" [6].
- Dify, Voiceflow and Relevance AI did not state approval mechanics on the pages we read; Relevance AI lists "agent evaluations" on its enterprise card [9].
The reason to insist on these features is not caution for its own sake. The human-factors literature shows that people over-rely on automated aids, and that complacency falls with accountability and training rather than with better interfaces A[10]. A meta-analysis of 106 human–AI experiments found that the combination performed worse than the better of the two alone on decision tasks, with gains only on creation tasks A[11]. An approval step is where the human stops being a rubber stamp. Our guide to managing AI agents turns that into a weekly routine.
Privacy and EU data residency
Only n8n states a region: "for hosted plans, data is stored within the EU, on servers located in Frankfurt, Germany" [5]. Lindy states "SOC 2 and GDPR compliant" without naming a region and offers HIPAA with a BAA on Enterprise [6]. Dify shows GDPR, SOC 2 and ISO 27001 badges and links a data-protection agreement, without a region [7]. Zapier, Make, Voiceflow and Relevance AI did not state residency, SOC 2 or GDPR on the pricing or agent pages we read [1][3][8][9]. For a European team that must tell its data-protection officer where personal data in an agent run is processed, that leaves n8n, or a self-hosted platform from our open-source agents comparison, as the only answers that do not require a sales call.
Security and permissions
A no-code agent is only as safe as the connections you give it. Four rules that hold across all seven tools: connect a service account rather than your own login where the app allows it; grant read scopes first and add write scopes one at a time; route every irreversible action (send, pay, delete, publish) through the approval feature; and keep the run logs for the first month, because you will need them to explain the one action nobody expected. Zapier, Make and n8n show step-level logs; Lindy shows task history; Relevance AI lists audit logs on Enterprise [9].
Who should not buy a no-code agent builder
- Anyone whose agent needs to run thousands of times a day. Above the n8n Pro tier the published prices end and the sales calls begin; a developer with a framework (see the open-source page) is cheaper.
- Anyone without a measurable task. The controlled evidence on AI assistance shows gains inside the model's capability and losses outside it, and the edge is hard to see in advance B[12]; if you cannot measure the agent's output, you cannot find the edge.
- Regulated data without a residency statement. Wait for the contract.
- Someone who wants a chat assistant, not an actor in their systems. The best AI agents for work covers the subscription agents inside Claude, ChatGPT and Microsoft Copilot, which may be all you need.
For the organisational view, costs and ownership across a company rather than one builder, see AI agents for business.
What to do next
Count runs and steps before you count features. If your agent will run a few hundred times a month inside apps you already connect, start with Zapier Agents on the free tier and watch the activity counter. If it will run thousands of times, or your data-protection officer has questions, take n8n Cloud Starter and switch on human approval for tool calls. If you think in Make scenarios, add Make AI Agents to the shortlist and read the price grid yourself, since we could not. Then run the Tool Finder with your run estimate and residency requirement: it returns the same dated prices and an automation ROI estimate for your volume.
Frequently asked questions
What is the cheapest no-code AI agent builder?
For a few hundred runs a month, Zapier Agents' free tier (400 activities) or Dify's free Sandbox (200 message credits) cost nothing. For thousands of runs, n8n Cloud is the cheapest published option: EUR 20 a month covers 2,500 executions and EUR 50 covers 10,000, with one execution per run regardless of steps. Prices read 2026-11-09.
How much does an AI agent cost per run?
It depends on the unit. Zapier counts activities (a five-step agent is about five per run, so the 1,500-activity paid tier covers about 300 runs); n8n counts one execution per run; Lindy counts credits from 2 to 2,500 per task; Dify counts message credits that vary by model. Make bills in credits per module action but its price grid did not render for us on 2026-11-09.
Can I build an AI agent without coding?
Yes. Zapier Agents, Make AI Agents, n8n, Lindy, Dify and Voiceflow all let you describe the agent in plain language, connect apps from a list and set rules or approval steps without code. The limits appear at volume, in error handling and in data residency, which is what this comparison measures.
Which no-code AI agent builder is GDPR-friendly for EU teams?
n8n is the only one of the seven whose pricing page states a region: hosted data is stored in the EU, in Frankfurt. Lindy states SOC 2 and GDPR compliance without a region; Dify shows GDPR, SOC 2 and ISO 27001 badges without a region. Zapier, Make, Voiceflow and Relevance AI did not state residency on the pages we read on 2026-11-09.
Do no-code AI agents work reliably?
They work as well as the approval step you build in. Zapier caps activities per run and asks before continuing, n8n offers human approval for tool calls, Make documents manual approvals and stop points, Lindy waits for approval on anything with outside impact. Human-factors research shows people under-check automated aids unless accountability is designed in, so use those features for anything irreversible.
Sources and evidence
- Zapier pricing (Agents activities per plan, Agents Pro price, task tiers) — Zapier (2026), vendor page A zapier.com/pricing (retrieved 2026-11-09)
- Zapier Agents (activity definition, per-run caps, knowledge sources) — Zapier (2026), vendor page A zapier.com/agents (retrieved 2026-11-09)
- Make pricing (FAQ text only; plan cards did not render) — Make (Celonis) (2026), vendor page B www.make.com/en/pricing (retrieved 2026-11-09)
- Make AI Agents — Make (Celonis) (2026), vendor page A www.make.com/en/ai-agents (retrieved 2026-11-09)
- n8n pricing (executions, human approval for tool calls, Frankfurt hosting) — n8n GmbH (2026), vendor page A n8n.io/pricing/ (retrieved 2026-11-09)
- Lindy pricing (credits, Team plan, approvals, SOC 2 and GDPR statement) — Lindy (2026), vendor page A www.lindy.ai/pricing (retrieved 2026-11-09)
- Dify pricing (Sandbox, Professional, Team; message credits) — LangGenius (2026), vendor page A dify.ai/pricing (retrieved 2026-11-09)
- Voiceflow pricing (agencies and businesses; no self-serve tiers shown) — Voiceflow (2026), vendor page B www.voiceflow.com/pricing (retrieved 2026-11-09)
- Relevance AI pricing (Enterprise card only) — Relevance AI (2026), vendor page B relevanceai.com/pricing (retrieved 2026-11-09)
- Complacency and Bias in Human Use of Automation: An Attentional Integration — Parasuraman R, Manzey DH (2010), systematic review A doi.org/10.1177/0018720810376055 (retrieved 2026-11-09)
- When combinations of humans and AI are useful: A systematic review and meta-analysis — Vaccaro M, Almaatouq A, Malone T (2024), meta-analysis A doi.org/10.1038/s41562-024-02024-1 (retrieved 2026-11-09)
- Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of Artificial Intelligence on Knowledge Worker Productivity and Quality — Dell'Acqua F, McFowland E, Mollick ER, et al. (2023), RCT B doi.org/10.2139/ssrn.4573321 (retrieved 2026-11-09)