Contents 13 sections
Make and n8n offer different kinds of free AI agents. Make hosts automations, while n8n offers a self-hosted route. Make lets users build and run automations without a credit card or time limit.1
This ranking compares hosted plans with open-source builders, including the costs that remain. Make’s free plan includes 1,000 credits per month2.
Comparison of free AI agents: table and ranking criteria
n8n, Make, Zapier Agents, Dify, Flowise, Langflow, and CrewAI differ most in where they run and what free access covers. Each tool is assessed on usage limits, separate model or API costs, and its cheapest paid monthly plan. n8n Cloud Starter includes 2,500 workflow executions per month; the self-hosted Community Edition is a different option3.
The table compares free access, usage limits, model or API costs, and confirmed monthly paid prices in ranking order.
Make Core costs $16/month with monthly billing and includes 10,000 credits4.
| Tool | Free option | Free limit | Model or API costs | Cheapest paid plan per month |
|---|---|---|---|---|
| 1n8n | Self-hosted Community Edition | 1,000 workflow executions (Cloud free trial; 14 days) | Separate for self-hosted use | $20/month (Starter, billed annually) |
| 2Make | Hosted free plan | Free plan: 1,000 credits/month | – | Core: $12/month billed annually for 10,000 credits/month |
| 3Zapier Agents | Hosted free access | 400 activities/month (Agents Free) | – | $33.33/month, billed annually ($400/year; Agents Pro) |
| 4Dify | Self-hosted option | 200 message credits (Sandbox plan) | Separate for self-hosted use | $59/workspace/month (Professional plan, monthly billing) |
| 5Flowise | Free plan: $0 per month | Free plan: 2 flows and assistants, 100 predictions/month, 5 MB storage | Separate for self-hosted use | – |
| 6Langflow | Self-hosted option | – | Separate for self-hosted use | – |
| 7CrewAI | Open-source framework | 50 workflow executions/month (Free plan) | Separate for self-hosted use | – |
Table data: 2,3,5,6,7,8
In this table, – means the cited figures do not establish a free agent limit, a separate model or API charge, or a paid price with a confirmed monthly billing period.
Zapier Agents has its own usage allowance. Zapier’s 100 tasks per month apply to Zaps, Tables, and Forms, not to Agents.5 Dify lists 5,000 message credits per month and lets users switch to their own API key afterward.9
These figures measure different things. Zapier Agents Free includes 400 agent activities per month, separate from the 100-task monthly Zaps allowance5.
n8n fits teams that want self-hosted workflow control and can cover infrastructure and model access. Make fits teams that would rather use hosted automation. Zapier Agents suits teams evaluating agents alongside Zaps, Tables, and Forms. Dify gives builders a choice between included model credits and their own API key.
Make’s hosted free plan includes 1,000 credits per month2.
Flowise still lists a $0-per-month plan, but its open-source project reached end of life on August 31, 2026. That status matters before starting a project. Langflow fits builders prepared to manage a self-hosted agent builder, while CrewAI suits developers seeking an open-source framework rather than a hosted free plan. Self-hosted software does not cover server or LLM API costs10.
n8n
4.1/5- Cheapest paid plan per month
- $20/month (Starter, billed annually)
- Free option
- Self-hosted Community Edition
- Free limit
- 1,000 workflow executions (Cloud free trial; 14 days)
At a glance
- n8n ranks first for teams that prefer self-hosted agent workflows to a hosted free-plan allowance.
- For example, n8n Cloud Starter includes 2,500 workflow executions per month, while Community Edition is self-hosted.
- n8n builds workflows: connected steps that move data between apps and agents.
n8n ranks first for teams that prefer self-hosted agent workflows to a hosted free-plan allowance. Its Community Edition is free software, but the team pays for its server and any LLM API use. Cloud plans are a separate choice with their own allowances.
For example, n8n Cloud Starter includes 2,500 workflow executions per month, while Community Edition is self-hosted3.
n8n: features confirmed by the vendor
n8n builds workflows: connected steps that move data between apps and agents. In the vendor’s refund example, a workflow looks up an order and requests a Slack decision about a refund or ticket.11 A person reviews that decision before the workflow acts. n8n documents nine human-approval channels for AI tool calls, including Slack and its built-in chat12.
n8n: free plan limits and real costs
n8n Community Edition is free to self-host; infrastructure and model calls still cost money. n8n lists its Cloud Assistant as a preview feature on Cloud plans except Enterprise. Starter gets 1,600 Assistant credits per month, while Pro gets 4,000 or 9,600 Assistant credits per month, depending on plan size.3
Those credits are not a Community Edition execution allowance. n8n Cloud Starter costs $24/month and includes 2,500 workflow executions13.
n8n lists saved-execution storage tiers of 2.5 GB, 25 GB, and 50 GB.3 Saved-execution storage retains records of workflow runs. Teams that need those records for troubleshooting should account for the storage tier.
n8n: who is it best for?
n8n suits teams able to manage deployment, connect apps, and pay for servers and models. It is a poor fit when free access must include managed hosting and LLM usage. For technically equipped teams, n8n offers control over self-hosted agent workflows.
Make
3.8/5- Cheapest paid plan per month
- Core: $12/month billed annually for 10,000 credits/month
- Free option
- Hosted free plan
- Free limit
- Free plan: 1,000 credits/month
At a glance
- Make hosts AI agent automations; n8n’s free Community Edition runs on user-managed infrastructure.
- Make connects to 3,500+ apps out of the box and supports custom integrations for services with an API.
- Make says it is GDPR compliant and holds SOC 2 Type II and SOC 3 certifications.
Make hosts AI agent automations; n8n’s free Community Edition runs on user-managed infrastructure. Make handles hosting and counts actions in a scenario as credits. Self-hosting instead puts server and model costs on the team.2
A standard Make module operation uses one credit, and some built-in AI modules charge for token usage too14.
Make: features confirmed by the vendor
Make connects to 3,500+ apps out of the box and supports custom integrations for services with an API.1 Its listed security features include encryption in transit and at rest, single sign-on, and access controls. Make says it is GDPR compliant and holds SOC 2 Type II and SOC 3 certifications.1 Those connections matter when an agent must act across an existing software stack.
Make: free plan limits and real costs
Make’s Free plan includes 1,000 credits per month, 1 MB of data storage, and 1 data store.2 A scenario is a Make automation. Its modules generally use credits when they perform actions, though the listed error-handler modules do not.2
The credit allowance, not just the number of scenarios, determines how much work fits into the free plan.
Make: who is it best for?
Make suits teams that want hosted automation and broad app connectivity without running a server. It is less suitable for teams that need self-hosted control or unlimited module actions. Make says its platform connects more than 3,000 native apps2.
- Small teams can compare their expected workflow activity with Make’s free-plan credit allowance before paying.
- Startups with fewer than 20 employees can check eligibility for Make’s Start-up Plan, which offers 50% off Business.2
Make is the strongest fit here for teams that value hosted, app-connected automation over self-hosted control. Its hosted free plan includes 1,000 credits per month2.
Zapier Agents
4.5/5- Cheapest paid plan per month
- $33.33/month, billed annually ($400/year; Agents Pro)
- Free option
- Hosted free access
- Free limit
- 400 activities/month (Agents Free)
At a glance
- Unlike self-hosted n8n, Zapier does not require the team to run a server.
- Zapier Agents can use company knowledge to work across 9,000+ apps.
- The free Agents plan includes 400 agent activities per month.
Zapier Agents hosts agents that act across business apps. Make centers its automation on scenarios, while Zapier emphasizes agents supplied with company knowledge. Unlike self-hosted n8n, Zapier does not require the team to run a server. Its general automation task allowance should not be confused with the Agents activity allowance.
Zapier Agents: features confirmed by the vendor
Zapier Agents can use company knowledge to work across 9,000+ apps.15 In the vendor’s sales-email example, an agent pulls website information to personalize a message.15 Zapier Copilot helps create specialized agents, but it is a distinct product for pricing purposes.15
The free Agents plan includes 400 agent activities per month5.
Zapier Agents: free plan limits and real costs
Zapier Agents pricing turns on automated behaviors, not the general Zapier task allowance described above. Zapier lists tiers with up to 400 and 1,500 automated behaviors per month. Those tier-specific limits are not interchangeable with Zapier tasks.5
New accounts can get a 14-day Professional trial, but that is temporary access rather than an ongoing free plan.5,16
Zapier Copilot is included in the Free plan, while $19.99 per month is the starting price for the Professional plan, not a standalone Copilot price; that is not a confirmed price for a paid Zapier Agents plan.5 Zapier says Agents do not currently support Enterprise account restrictions on apps and actions.5
Zapier Agents: who is it best for?
Zapier Agents suits teams that want agents to use company knowledge across connected apps. It does not suit organizations that require Enterprise app and action restrictions to govern agents. Agents Free allows 400 activities per month, Agents Pro allows 1,500, and Enterprise limits are custom5.
- App-connected teams can use Zapier Agents when an agent needs to act across their business software.
- Enterprise administrators should treat the restriction gap as a blocker when those controls are mandatory.
Zapier Agents is the best fit for teams seeking hosted, broadly connected agents that do not depend on those Enterprise restrictions.
Dify
3.6/5- Cheapest paid plan per month
- $59/workspace/month (Professional plan, monthly billing)
- Free option
- Self-hosted option
- Free limit
- 200 message credits (Sandbox plan)
At a glance
- Dify lists a 50-document knowledge quota and 3,000 trigger events; these track different activities, not a shared agent-execution limit.
- Dify lists a 17% discount for annual billing compared with monthly billing.
Dify builds agents and AI apps through a self-hosted Community Edition or managed Cloud option.17 Unlike a hosted-only platform, it lets teams choose who runs the infrastructure. The self-hosted software is free, though the server and model access are not.
Dify: features confirmed by the vendor
Dify supports building and deploying AI apps in Cloud or on private infrastructure.9 A knowledge data source lets an app use imported documents. Plugins, schedules, and webhooks can start workflows automatically.9
Dify lists a 50-document knowledge quota and 3,000 trigger events; these track different activities, not a shared agent-execution limit.9
Dify: free plan limits and real costs
Dify Community Edition is source-available and can be self-hosted with Docker. Its software access does not include paid hosting or LLM API usage.17 Dify Cloud offers managed hosting and message credits for trying models before switching to a user-supplied API key.9
Credit use depends on the selected model, so a message credit does not guarantee a set number of agent runs.9
Dify lists a 17% discount for annual billing compared with monthly billing.9 That payment commitment does not change what the free self-hosted option includes.
Dify: who is it best for?
Dify suits builders creating document-backed agents who want a choice of managed or private deployment. It is not a fit for teams seeking free managed hosting with unlimited model use.
- Self-hosting teams can use Dify Community Edition if they can operate Docker and budget for servers and models.
- Cloud-focused teams can choose Dify’s managed option to avoid operating the deployment themselves.
Dify is the best fit for teams that value deployment choice and document-backed AI apps over a single hosted automation workflow.
Flowise
4.3/5- Free option
- Free plan: $0 per month
- Free limit
- Free plan: 2 flows and assistants, 100 predictions/month, 5 MB storage
At a glance
- The vendor lists connections to 100+ sources, tools, vector databases, and memory components.
- Its paid plan for individuals and small teams costs $35 per month and includes 10,000 predictions per month.
Flowise is an open-source builder for AI agents and LLM workflows, with cloud and on-premises deployment options.7,18 It offers more deployment control than a hosted automation platform, but self-hosting leaves infrastructure and model costs with the team. Paid-tier usage is measured in predictions, not automation credits.
Flowise: features confirmed by the vendor
Flowise supplies modular building blocks for workflows and autonomous agents.7 Its Agentflow builder supports chat assistants, single-agent and multi-agent systems, and workflow orchestration. The vendor lists connections to 100+ sources, tools, vector databases, and memory components.18 That makes Flowise a builder for customized agent systems rather than only preset app automations.
Flowise: free plan limits and real costs
Flowise lists a free tier; teams that self-host its open-source software pay their own server and LLM API costs.7,18
Its paid plan for individuals and small teams costs $35 per month and includes 10,000 predictions per month.7 A prediction is an AI application response. The paid prediction allowance is not a free-tier limit.
Flowise: who is it best for?
Flowise suits builders who want modular agents and deployment control, provided they account for its end-of-life status. It is not a fit for teams expecting self-hosted software to include free server capacity and model calls.
Flowise is the best fit for developers who prioritize customizable agent workflows over ready-made hosted automation. Its sunset status remains a consideration.
Langflow
4.3/5- Free option
- Self-hosted option
At a glance
- Langflow is a free, self-hosted builder for creating and serving application workflows.
- Langflow creates and serves flows, functional representations of application workflows.
Langflow is a free, self-hosted builder for creating and serving application workflows. It offers an enterprise-grade cloud deployment option too, though this ranking considers its self-hosted route as the free option.19 Teams remain responsible for server and LLM API costs.
Langflow: features confirmed by the vendor
Langflow creates and serves flows, functional representations of application workflows.20 Builders can deploy a defined flow rather than start with a preset agent task. Langflow’s cloud platform offers a separate way to deploy and scale those workflows.19
Langflow: free plan limits and real costs
Langflow’s self-hosted builder does not require a software subscription. It does require server capacity and model API usage, while cloud deployment is a separate option.19
Teams should include hosting and model calls when estimating what a deployed flow will cost to run.
Langflow: who is it best for?
Langflow suits builders who want to define and serve flows on infrastructure they manage. It is not a fit when free access must include managed hosting or model usage.
Langflow is the best fit for technically equipped teams that want control over application workflows.
CrewAI
4.1/5- Free option
- Open-source framework
- Free limit
- 50 workflow executions/month (Free plan)
At a glance
- Its vendor describes a prompt-and-canvas builder guided by 700,000 agent workflow patterns.
- The vendor reports a customer use case with 3,000+ leads enriched per month.
CrewAI is a free, open-source framework for collaborative AI agents. Unlike a hosted free plan, it gives developers software to run on their own infrastructure. Server and LLM API usage remain separate costs. CrewAI supports deployment in its cloud or in a team’s environment8.
CrewAI: features confirmed by the vendor
CrewAI builds agents that work together on operational tasks. Its vendor describes a prompt-and-canvas builder guided by 700,000 agent workflow patterns. Tracing records LLM calls, tool calls, memory reads, and their costs.21 Connected apps, including Gmail, Slack, and Salesforce, can trigger workflows22.
CrewAI: free plan limits and real costs
CrewAI’s open-source framework is the free option here, not an allowance of hosted executions or model credits. Teams running agents on their own infrastructure pay hosting and LLM API charges. CrewAI Cloud is a separate deployment choice.8
Tracing can help attribute those charges to agent activity, but it does not eliminate them.21
CrewAI: who is it best for?
CrewAI suits developers building agents that coordinate work across connected systems. The vendor reports a customer use case with 3,000+ leads enriched per month. That result is not a free-plan allowance.21 CrewAI is not a fit for someone who needs free hosted agents with infrastructure and model usage included.
CrewAI is the best fit for developers who want collaborative agent workflows and control over deployment.
Methodology and limits of this comparison
Self-hosted agent builders and hosted free plans differ in what stays free, who pays operating costs, and what limits usage. The ranking follows the specified product order rather than a performance score.
As of November 2026, this assessment draws on vendor pages, pricing pages, and documentation, not hands-on performance testing.
The table compares self-hosted and hosted free options by access, operating costs, and usage constraints.
| Free option | What is free | Who pays operating costs | Usage constraint |
|---|---|---|---|
| Self-hosted software | Software access | User pays for servers and model APIs | Infrastructure and model usage |
| Hosted free plan | Vendor-hosted access | Vendor covers included hosting; user pays for usage outside the plan | Plan-specific credits, tasks, or executions |
Vendor-stated allowances use different units. A task, credit, prediction, and agent behavior do not represent the same amount of work. Customer case studies are not plan limits, and trials do not establish ongoing free access.
The useful comparison is what the intended workflow costs and where its usage limit falls.
Self-hosted free vs hosted free plan
Self-hosted free software gives teams deployment control while leaving servers and model APIs to them. A hosted free plan removes server setup but imposes the vendor’s usage allowance. n8n Community Edition is free to self-host without a license key.23 Langflow offers both self-deployment and a free cloud account, so self-hosting is not its only deployment choice.19
The table compares setup effort, costs beyond free access, and usage conditions for both approaches.
| Option | Setup effort | Costs outside free access | Usage limit or condition |
|---|---|---|---|
| n8n Community Edition | Manage the server | Server and model API usage | Runs depend on the user’s infrastructure |
| n8n Cloud trial | Use vendor hosting | Paid access after the trial | 5 concurrent executions and 800 Assistant credits. |
| Zapier Free plan | Use vendor hosting | Usage beyond the plan | 15-minute polling interval for checking for new events. |
| Flowise Free plan | Use vendor hosting | Usage beyond the plan | 2 flows and assistants, 100 predictions per month, and 5 MB storage. |
| Langflow free cloud account | Use vendor hosting | Usage outside the account’s included access | Account-based cloud access. |
Table data: 3,5,7,19,24
Its Business trial lasts 14 days.3 Neither provides the continuing free software access of Community Edition. Flowise says it is being sunset, which matters before building around its hosted free plan.7
Self-hosted software fits teams that need deployment control and can operate the infrastructure. Hosted free plans fit teams that would rather avoid server setup and work within vendor limits.
Hidden costs of “free” agents
n8n, Dify, Make, and CrewAI can be used without a software charge. A useful cost comparison still includes model calls, hosting, usage limits, and connected apps. Paid n8n plans meter qualifying production workflow executions rather than individual steps; n8n says its self-hosted Business plan counts a production run whether it succeeds or fails.3
Its Cloud trial includes 1,000 executions and a 180-second execution timeout; it is not an ongoing free plan.3,25
The checklist covers costs and restrictions to consider before an agent runs beyond its free allowance.
- LLM API usage needs its own budget when a team supplies the model key. Dify lets users switch to their own API key after using the included message credits.9
- Server costs belong in a self-hosted agent’s total cost. Add monthly infrastructure charges to model usage even when the software is free.
- Execution limits must suit the intended workload. Dify lists an entry-level limit of 10 knowledge requests per minute and 2 triggers per workflow.6,9
- Integrations and premium apps may carry separate restrictions. Check whether each connected service permits the intended agent action under the team’s existing plan.
- Commercial rights and data use need review before an agent handles client information. Check the terms for outputs, uploaded knowledge, and data sent to model providers.
CrewAI’s free Basic plan includes a visual editor, AI copilot, GitHub integration, tracing, guardrails, and human-in-the-loop input.8 Those building features do not pay for the infrastructure or model usage behind a deployed agent.
The practical total includes the plan charge, server costs, model APIs, and connected apps.
When to upgrade
Make, Dify, CrewAI, and Zapier warrant an upgrade when a recurring workflow hits a limit or needs a paid feature. Simply having a free option is not a reason to upgrade. Make can pause a scenario when credits run out, while CrewAI limits Basic-plan executions and automations.2,8
- Make pauses scenarios when credits run out and resumes them when credits are added. Review credit use by action-performing modules; routers and error handlers do not consume credits.2
- Dify’s entry-level tier includes 200 model-trial message credits, 50 MB of knowledge storage, and 30 days of log history. Consider upgrading when model use, documents, or troubleshooting needs exceed those allowances.6,9
- CrewAI Basic includes 50 executions per month and 2 agentic workflow automations. Enterprise merits consideration when the team needs governance features or deployment on customer-controlled infrastructure.8
- Zapier MCP uses 2 tasks per tool call from the account’s task quota. Estimate tool calls before upgrading an automation plan; the Free plan also limits daily Copilot messages.5
Upgrade decisions should start with the allowance the workflow actually uses. Then compare the relevant paid plan with reducing usage or self-hosting. Flowise’s paid prediction allowance appears above; the other limits help identify when paid access is worth considering. Credits, tasks, and executions are not equivalent.
Sources
- ↑AI Workflow Automation Software & Tools | Make https://www.make.com/
- ↑Pricing & Subscription Packages | Make https://www.make.com/pricing
- ↑n8n Plans and Pricing – n8n.io https://n8n.io/pricing/
- ↑StackOnSale, Make vs. Zapier pricing https://stackonsale.com/guides/make-vs-zapier
- ↑Plans & Pricing | Zapier https://zapier.com/pricing
- ↑Dify Cloud pricing. ([dify.ai](dify.ai https://dify.ai/pricing/dify-cloud
- ↑Flowise – Build AI Agents, Visually https://flowiseai.com/
- ↑Pricing | CrewAI https://crewai.com/pricing
- ↑Dify Pricing – There’s Always a Plan for You https://dify.ai/pricing
- ↑Flowise sunset announcement; Langflow deployment documentation; CrewAI official repository. ([flowiseai.com](flowiseai.com https://flowiseai.com/sunset
- ↑AI Workflow Automation Platform – n8n https://n8n.io/
- ↑n8n Docs, Human-in-the-loop for tools https://docs.n8n.io/build/integrate-ai/ai-examples/human-in-the-loop-for-tools.md
- ↑n8n, n8n vs. Activepieces https://n8n.io/vs/activepieces/
- ↑Make Help Center, Credits https://help.make.com/credits
- ↑Build AI teammates with Zapier Agents https://zapier.com/agents
- ↑Zapier Agents usage documentation and Zapier Free-plan documentation. ([help.zapier.com](help.zapier.com https://help.zapier.com/hc/en-us/articles/26559132765325-How-is-Zapier-Agents-usage-measured
- ↑Dify – AI Agent Management Platform That Puts You in Control https://dify.ai/
- ↑Introduction | FlowiseAI https://docs.flowiseai.com/
- ↑Langflow | Low-code AI builder for agentic and RAG applications https://www.langflow.org/
- ↑What is Langflow? | Langflow Documentation https://docs.langflow.org/
- ↑CrewAI https://crewai.com/
- ↑CrewAI Documentation – CrewAI https://docs.crewai.com/
- ↑Host n8n | Deploy | n8n Docs https://docs.n8n.io/deploy/host-n8n
- ↑Zapier Free-plan documentation. ([help.zapier.com](help.zapier.com https://help.zapier.com/hc/en-us/articles/32337438839565-What-s-included-in-Zapier-s-Free-plan
- ↑docs.n8n.io https://docs.n8n.io/build/understand-workflows/understand-executions
Prices and terms checked at the source: November 2026.