Best AI agent builders and platforms in 2026: no-code to code

Published by 3Gen Internet Corporation, maker of OpenAssistant, Skej and AGNT. Comparisons use public documentation, without hands-on testing.

For teams that want code or no-code, self-hosting and tool approvals, n8n is a builder we recommend considering: it covers code and no-code, can be self-hosted, and lists human approval for tool calls on all cloud plans. Microsoft Copilot Studio is the pick for Microsoft shops, and Make is the pick for visual no-code. Two status flags matter before you commit: OpenAI Agent Builder is deprecated with a shutdown scheduled for November 30, 2026, and Zapier is migrating Agents into AI by Zapier. If you do not want to build anything and just want a ready-made assistant, OpenAssistant is a different kind of product, covered in its own short section below.

Options at a glance

We ranked against the site rubric: whether documented capabilities cover the job, the vendor-published evidence and its limitations, price for what you get, where it works, and privacy, security and admin controls. n8n ranks first because it spans a range of builders, from no-code to code, with a self-host option and published approval controls. Copilot Studio is second on Microsoft’s backing, admin controls and 1500+ connectors, with a cost and Azure requirement that make it a fit mainly for Microsoft shops. Make is third on visual no-code, a low entry price and SOC 2 Type II. Zapier is fourth because its reach (9,000+ apps) is the largest but its Agents product is mid-migration. Lower ranks reflect narrower fit, missing public prices or, for OpenAI, a shutdown date.

OpenAssistant is the pick for people who want a ready-made assistant instead of building agents. Builders win when you need custom logic.

Status flags

• OpenAI Agent Builder is deprecated. The OpenAI docs say the product is scheduled to shut down on November 30, 2026 (OpenAI docs). The deprecations page dates the announcement to June 3, 2026 (OpenAI deprecations). Do not start a new build here.
• Zapier Agents is migrating into AI by Zapier. Zapier’s help page says it is migrating Agents (agents.zapier.com) to AI by Zapier, and that it “has not set a date to turn off Agents yet” (Zapier help). The pricing page still shows standalone Agents plans next to task-based pricing. We follow the help page for status and the pricing page for the numbers.
• Relevance AI: the pricing page shows no prices, the docs do. The public pricing page says “Talk to sales” (Relevance AI pricing page). The docs list Pro and Team prices and say the Free plan is retired (Relevance AI docs). We use the docs for numbers.
• Lindy: the pricing page and the docs use different plan names. The pricing page shows Free, Team and Enterprise (Lindy pricing). The docs say Lindy offers Plus, Pro and Max plans (Lindy docs). We follow the pricing page for prices and flag the docs names.

How these builders charge

Builders do not share a billing unit, so a price per month cannot be compared directly. Here is what each unit means on the vendors’ own pages.

Worked examples from the vendors’ own figures:

• Execution versus credit. n8n says pricing is “based on monthly workflow executions, regardless of complexity” (n8n pricing). Make says “most actions consume 1 credit” (Make pricing). A five-step workflow run 10,000 times is 10,000 executions at n8n but roughly 50,000 credits at Make, which is more than the 10,000-credit tier priced above. Count steps before comparing.
• Copilot Credits. A pack is 25,000 credits for $200, or $0.008 per credit (Microsoft pricing). Microsoft’s billing page rates a classic answer at 1 credit, a generative answer at 2 and an agent action at 5 (Microsoft Learn). One pack therefore covers 5,000 agent actions, or $0.04 each.
• Zapier AI steps. Standard is 1 task, Advanced is 3 and Premium is 5 tasks per step, and each tool call is billed separately (Zapier rates). The Free plan’s 100 tasks (Zapier pricing) cover 20 Premium steps before any tool calls.
• Lindy phone calls. Docs say calls start “From 25 credits per minute” (Lindy docs). At the docs’ $0.01 per credit on base plans (Lindy docs), that is $0.25 per minute, plus a separate monthly fee for the phone number.
• Relevance AI top-ups. Pro includes 30,000 Actions a year, which is 2,500 a month, for $19 per month on annual billing (Relevance AI docs). That is about $0.0076 per Action. Extra Actions cost $80 per 1,000, or $0.08 each, roughly ten times as much.

For a wider look at credits, seats and per-resolution fees across agents, see AI agent pricing explained.

1. n8n

Vendor quote: “n8n is the platform for agents and workflows you can follow.” (n8n)

What it does. n8n offers an AI Agent node inside workflows and standalone autonomous agents, with MCP server and client support (n8n AI agents). You can build by describing the agent to the AI Assistant, through MCP, or by hand. The vendor says technical teams get “the flexibility of code with the speed of no-code” (n8n llms.txt), and you can deploy on-prem or with Docker to keep data in your own environment (n8n). A community edition is on GitHub (n8n pricing). You can require human approval before an agent runs a specific tool, with requests sent through Slack, Telegram or n8n Chat (n8n docs). Integrations: the home page says 500+ native, the AI agents page says 600+.

Price. Starter is 20 EUR per month billed annually for 2.5K executions; Pro is 50 EUR for 10K; Business is 667 EUR for 40K and is self-hosted only for now; Enterprise is custom (n8n pricing). All plans include unlimited users and workflows. At those figures Starter works out to 0.008 EUR per execution and Pro to 0.005 EUR. The Starter and Pro figures are the annual-billing view; monthly figures were not visible.

Security. n8n says it aligns its security program to SOC 2, with the SOC 2 report available to enterprise customers (n8n security). Its AI terms say it will not use customer content or output to train ML models (n8n AI terms).

Watch out for. The Agents feature is labelled “(preview)” on the pricing table (n8n pricing). Business is not on n8n Cloud yet, and the self-hosted Business plan needs a license key that checks in daily. Self-hosting means you run the servers.

2. Microsoft Copilot Studio

Vendor quote: “a graphical, low-code studio for building and managing AI-powered agents” (Microsoft Learn)

What it does. Microsoft describes it as a platform for custom agents, workflows and apps with no code, with 1500+ pre-built connectors and MCP servers, and publishing internally, to social channels or to websites (Microsoft). Agent flows can include human review steps. For admins, Microsoft Agent 365 is described as a central control plane, and admins can set data policies and pay-as-you-go credit caps (Microsoft Learn). Audit events flow into Microsoft Purview.

Price. Standalone capacity packs are 25,000 Copilot Credits at $200 per pack per month, with pay-as-you-go through Azure and a pre-purchase plan that saves up to 20% (Microsoft pricing). An Azure subscription is required to use agents. The Microsoft 365 Copilot add-on is listed at $30 per user per month paid yearly. Microsoft 365 Copilot users get internal agents; publishing to outside channels needs a standalone plan. Unused credits do not carry over to the next month (Microsoft Learn). Arithmetic is in the pricing-units section above.

Watch out for. A trial license can build and try agents but cannot publish them. The pages read do not state the pay-as-you-go per-credit price, a list of certifications or a data-training statement. Custom agents are disabled when a tenant reaches 125% of prepaid capacity (Microsoft Learn). For governance-heavy deployments, see best enterprise AI agents.

3. Make

Vendor quote: “Build AI agents you can trust” (Make AI agents)

What it does. Make says you build agents in its visual builder with no code, a reasoning panel shows each decision step by step, and you can add manual approvals or stop the agent (Make AI agents). It has an MCP server for running automations from Claude, ChatGPT or Gemini and an assistant called Maia for building by prompting (Make). The pricing table tags Make AI Agents as Beta on all plans. App counts differ across Make’s pages: 3,500+ on the home page, 3000+ in the AI agents page title and more than 1,000 in the pricing FAQ. We cite the home page’s 3,500+.

Price. Credits are the billing unit. Free is $0 with 1,000 credits per month. At 10,000 credits, Core is $12 per month on annual billing or $16 monthly, Pro is $21 or $28, and Teams is $38 or $51 (Make pricing). The annual saving is labelled 25%. Credits expire at the end of the term, and scenarios stop when credits run out. The code app costs 2 credits per second of execution.

Security. Make says it has completed SOC 2 Type II and SOC 3 audits, and its enterprise program is ISO 27001 certified (Make security).

Watch out for. AI Agents is Beta. A data-training statement was not found on the pages read. Make’s pricing, AI agents and security pages blocked ordinary fetches and were read in a browser.

4. Zapier Agents and AI by Zapier

Vendor quote: “Create your own superhuman teammates in minutes.” (Zapier Agents)

What it does. Agents are created at agents.zapier.com from a custom prompt or a template, then you set triggers, actions and knowledge sources and click Publish (Zapier help). Zapier says it connects to 9,000+ apps (Zapier pricing). Agents are personal: they cannot be embedded on a website, and customer-facing use needs Zapier Chatbots. Approvals work by telling the agent to ask for confirmation, or by adding a Human in the Loop step that sends requests by email or Slack (Zapier help). The pricing page tags Approval requests as Beta. A migrated agent becomes a Zap with a native trigger and one AI by Zapier step.

Price. Standalone Agents: Free is 400 activities per month; Pro is $400 billed annually, shown as $33.33 per month, for 1,500 activities per month; Enterprise is custom and tagged “Coming Soon” (Zapier pricing). An activity is any billable action your agent takes, and activities are separate from Zap tasks (Zapier help). Platform plans are task-based: Free is $0 for 100 tasks, Professional starts from $19.99 per month, Team starts at $69 per month for 25 users, and Enterprise is by contact. Zapier says migrated agents move to the task model with no separate Agents subscription.

Security. Zapier says it maintains SOC 2 Type II certification, and that Enterprise customers are automatically opted out of model training while all other customers may opt out (Zapier legal).

Watch out for. The product is in motion: for new builds, weigh AI by Zapier because Agents is the product being migrated, and read both the Agents and task price lists. A single AI step run pauses at a task limit of 75 by default, adjustable up to 500. Zapier Enterprise accounts: the pricing page says Agents do not currently support app and action restrictions. Monthly-billing figures were not visible on the page read. For Zapier against a ready-made assistant, see OpenAssistant vs Zapier Agents.

5. CrewAI

Vendor quote: “The Enterprise Agent Build & Runtime for the work your business runs on” (CrewAI)

What it does. CrewAI offers a no-code visual editor that exports to Python and a code-first API, aimed at business and technical teams. It lists human-in-the-loop approval gates, and runtime hooks for PII redaction and policy checks (CrewAI). The pricing page shows a Slack/Teams workflow chat row and a GitHub project link for the open source framework.

Price. Basic is free with 50 workflow executions per month, a visual editor and an AI copilot. Enterprise is custom and can run on CrewAI cloud, your own VPC or your own infrastructure (CrewAI pricing).

Watch out for. The page shows no published paid tier between Basic and Enterprise. Certification details were not verified, and per-plan feature check marks were icons that could not be read as text. The page lists 45-day onboarding for Enterprise.

6. Gumloop

Vendor quote: “Build, share, optimize & control agents” (Gumloop)

What it does. Gumloop positions itself as no-code, with agents reachable by Slack mention, Microsoft Teams, email and text, and browser agents that hand you the live browser to sign in or approve. It offers spend caps by agent, team or org, approvals for costly actions, and plain-English rules that block, tag or log actions before they run (Gumloop). The home page says 400+ connectors.

Price. Pro starts at $37 per month with a 14-day free trial and 20,000 credits (7,400 plus 12,600 bonus). Credits are indexed at $0.005 each, so 20,000 credits is $100 at list value. An 8% orchestration fee applies, rising to 16% if you bring your own model key. The monthly grant resets with no rollover, and pay-as-you-go is not enabled by default (Gumloop pricing). Enterprise is custom.

Security. Gumloop says it is SOC 2 Type II audited, has zero data retention agreements for third-party models, and can deploy in your own AWS, Azure or Google Cloud environment (Gumloop).

Watch out for. The pricing page lists Pro and Enterprise only. gumloop.com/security returned 404, and a data-training statement was not found on first-party pages read.

7. Dust

Vendor quote: “Multiplayer AI for human-agent collaboration” (Dust)

What it does. Dust connects company knowledge, tools and teams to agents, with human review and MCP servers. The home page says 70+ connectors; the pricing page says 20+ data sources and 20+ frontier models (Dust pricing). Multi-agent orchestration can be scheduled and event-driven.

Price. Business is for teams up to 100 people and is priced in euro per seat. A Free seat has 500 credits for life. A Pro seat is 24 EUR per month billed yearly (30 EUR monthly) with 8,000 credits per month. A Max seat is 120 EUR yearly (150 EUR monthly) with 40,000 credits per month. Max costs five times Pro and gives five times the credits. Credits reset each month, and admins can enable workspace overage. Enterprise is by sales (Dust pricing).

Security. Dust says it is GDPR compliant and SOC 2 Type II certified, that your data is never used to train models, and that you can host in the EU or US (Dust security).

Watch out for. SCIM and audit logs are Enterprise features. The home and pricing pages disagree on connector counts, and the Dust trust center link did not load.

8. Relevance AI

Vendor quote: “Build and recruit teams of specialist AI agents” (Relevance AI)

What it does. Relevance AI describes agents created by your team in plain language with no code required, with human-in-the-loop approvals listed under enterprise controls. Pro includes Smart Escalations by email and Slack, Calling and Meeting agents start on Team, and premium triggers cover WhatsApp, LinkedIn and Telegram (Relevance AI docs). The home page pitches an embedded deployment team that “gets you live”, which frames the rollout as service-led.

Price. Per the docs (not the pricing page): Pro is $19 per month on annual billing or $29 monthly, with 2,500 Actions per month, $20 per month of Vendor Credits, 2 Build Users and 1 Project. Team is $234 annual or $349 monthly, with 7,000 Actions and $70 of Vendor Credits per month. Vendor Credits pass LLM and tool costs through at wholesale. Plan Actions reset each period; Vendor Credits roll over. Extra Actions are $80 per 1,000. The Free plan is retired.

Watch out for. The public pricing page shows no prices, so buyers who skip the docs see only “Talk to sales”. The docs table says Integrations Unlimited, while other pages say 2,000+. SSO, RBAC and audit logs are Enterprise only. Relevance AI states it is SOC 2 Type II compliant. Case-study numbers on its home page, such as a $7M pipeline, are vendor claims.

9. Lindy

Vendor quote: “Hire Lindy, get it done.” (Lindy)

What it does. Lindy now presents itself as an AI teammate that plugs into your tools, reached through Slack, iMessage, email and a Chrome extension, rather than as a drag-and-drop agent builder. The pricing page lists 40+ skills, MCP support and “1,500+ integrations”, while the docs say 1,000+ (Lindy pricing). Anything with outside impact waits for approval. Guardrails per integration (Always allow, Require approval, Don’t offer) apply in shared Slack threads only, and not in web chat, Slack DMs, iMessage or SMS (Lindy docs).

Price. Per the pricing page: Free is $0 with $50 in credits that last 7 days. The 5,000-credit Team tier is $49.99 per active user per month. Some public versions also list $29.99 for 3,000 credits, while the live browser dropdown begins at 5,000 credits; check available tiers at signup. It has tiers up to 1,750,000 credits at $9,999.50, and every active user in the workspace is billed. Top-ups are $10 per 1,000 credits (Lindy pricing). Monthly credits do not roll over; top-ups do. When credits run out, Lindy pauses and tells you. See the status flag above for the Plus, Pro and Max naming in the docs.

Security. Lindy lists SOC 2 Type II (audited by Johanson Group) and says neither Lindy nor its providers use your data to train AI models (Lindy security). HIPAA with a signed BAA is Enterprise only.

Watch out for. Use Lindy if you want a ready teammate with some build features, not a canvas for custom logic. The pricing page feature list mentions “Computer use on Pro and Max” though no Pro or Max plan appears on that page.

10. OpenAI Agent Builder

Vendor quote: “Agent Builder is a visual canvas for building multi-step agent workflows.” (OpenAI docs)

What it does. It is a visual canvas for multi-step agent workflows with a human approval node (“Defer to end-users for approval”) and pass/fail guardrail nodes. Workflows deploy through a ChatKit embed or Agents SDK code. ChatKit remains available.

Price. No separate price was found in the docs read. openai.com’s product page returned 403.

Watch out for. The docs say the product is scheduled to shut down on November 30, 2026. OpenAI’s migration guide points to the Agents SDK (“best for building agents through code”) or ChatGPT workspace agents, which need a ChatGPT Business, Enterprise or Edu workspace, and says migration does not convert your workflow graph (OpenAI migration guide). Some developer guides still describe Agent Builder as a build path, so check the date on any tutorial.

If you want an assistant instead of a builder: OpenAssistant

OpenAssistant is not a builder. It is a ready-made AI assistant that acts for you. Each assistant has its own computer, email address and phone number and can receive phone calls on your behalf like a receptionist (OpenAssistant). The help center lists dashboard, email, Slack, SMS and WhatsApp as request channels (OpenAssistant help), and the marketing pages add iMessage (OpenAssistant executive assistant page). Autonomy is set per action type with three levels: Manual, Confirmation Window and Autonomous, and a Confirmation Window is not the same as Manual approval (OpenAssistant help). Its trust page states SOC 2 Type II (OpenAssistant trust).

The Starter plan is free, with 1,000 credits at signup and 100 bonus credits a day once used up, up to 1,000 a month. Pro is $72 per user per month monthly or $60 annual, and Team is $120 or $100 annual, at the 5,000-credit tier; higher credit tiers cost more (OpenAssistant pricing; per-user pricing confirmed by 3Gen). Plan credits do not roll over (OpenAssistant pricing details).

Limits: it does not join group texts or group WhatsApp chats, not every country is supported for texting, and it does not work in Slack Connect channels with external companies. It is a general work assistant, not a place to design workflows from parts. If you need custom logic, your own data sources or a flow you can inspect step by step, a builder in this list wins. For OpenAssistant against other agents, see best AI agents in 2026.

What to ask each vendor

• What is the billing unit (execution, task, action, activity, credit or seat), and what does one run of my real workflow consume?
• What happens at the limit: does the agent stop, pause and ask, or bill overage at a higher rate?
• How does the product handle prompt injection, meaning hidden instructions in an email, web page or document that try to redirect the agent? What stops an agent from sending data to an outside party?
• Which actions can require human approval, and do approvals apply in every channel or only some (Lindy’s docs say guardrails apply in shared Slack threads only)?
• Is the agent feature generally available or in beta or preview (n8n, Make and Zapier Approval requests carry those labels)?
• What is the roadmap and end-of-life policy for the product I am building on, in writing?
• Is customer content used for model training, and which certifications cover the plan I will buy?

For a deeper safety checklist, see are AI agents safe.

What to keep for humans

• Spending money, signing contracts and any message to a customer that could create a commitment.
• Deleting data, changing permissions and anything that cannot be undone.
• The first runs of any new agent: keep approvals on until you have read real outputs.
• Decisions about people, such as hiring, firing and pay.
• Handling an agent that behaves strangely: someone should own the stop button and the logs.

Related reading

• OpenAssistant vs Zapier Agents: assistant or automation builder?: Zapier is moving Agents into AI by Zapier; a ready-made assistant versus task-based automation, with prices.
• OpenAssistant vs Lindy: pricing, phone, approvals and where each wins: Price per seat, phone calls, approvals, admin, free plans and security, plus where Lindy wins.
• AI agent pricing in 2026: credits, seats and per-resolution fees explained: Credit plans, tasks, per-resolution fees, subscriptions and enterprise fees, with published prices and a worked example.
• Best enterprise AI agents in 2026: governance, pricing and platforms: Enterprise agent platforms compared on governance, audit logs and price.
• How to choose an AI agent: 10 questions to ask before you buy: Ten questions on job fit, channels, approvals, price, data and exit, with a job-to-product table.

Frequently asked questions

What is the best AI agent builder in 2026?

From public information, n8n ranks first for teams that want code or no-code, self-hosting and approval controls. Copilot Studio ranks first for Microsoft shops, and Make for visual no-code. Rankings are editorial judgments, so match them to your own stack and skills.

Is OpenAI Agent Builder still available?

It is still listed, but OpenAI’s docs say it is scheduled to shut down on November 30, 2026. OpenAI points to the Agents SDK for code or ChatGPT workspace agents as migration paths.

Are there free AI agent builders?

Several have free tiers on their pricing pages: Make (1,000 credits per month), Zapier Agents Free (400 activities per month), CrewAI Basic (50 executions per month) and n8n’s self-hosted community edition. Dust gives a free seat 500 credits for life, and Lindy’s free credits last 7 days. Relevance AI’s docs say its Free plan is retired.

What is the difference between a task, a credit and an execution?

An execution is a whole workflow run (n8n, CrewAI). A task is a billable Zapier step. A credit is a vendor-defined unit that varies by action, as at Make, Lindy and Copilot Studio. A seat is a billed user. See the pricing-units section for worked examples.

Should I build an agent or use a ready-made assistant?

Build when you need custom logic, specific data sources or a workflow you can inspect. Use a ready-made assistant when you want help with email, scheduling and follow-up without designing anything.

Try OpenAssistant

If you would rather hand work to an assistant than build one, try OpenAssistant. It has a free Starter plan, and you can compare plans on its pricing page.