MMake AI Agents

Make AI Agents Put Every Decision on the Workflow Canvas

More than 3,000 apps sit behind Make’s next-generation AI Agents, which the company released on February 11, 2026 after a waitlist that opened in October 2025. Each agent is built, run and debugged on the same canvas as a Make scenario, and a Reasoning Panel shows in real time which tool it chose and why.

Make AI Agents suit operations, marketing and support teams that already automate in Make and want parts of a workflow to make judgment calls. The Free plan includes 1,000 credits a month, and paid plans start at $9. AI steps can use more credits than ordinary actions, so usage needs watching. Make AI Agents score 9.1/10 on our scale.

Make’s Agent Release, App Count and Review Figures

Make’s launch post, its pricing and comparison pages and G2 supply these facts.

The app count is the figure that matters most for an agent. An agent is only as useful as the tools it can call, and Make’s library gives it thousands without custom connectors. Make’s own pages cite 3,000+ in some places and 3,500+ in others, so the exact number depends on how modules are counted.

Make also stresses that the agents mix AI decisions with deterministic steps. That design choice is aimed at teams that cannot hand a whole process to an AI and need most steps to behave the same way every time.

The waitlist timing also tells buyers something. Make announced the new agents in October 2025 and opened them in February 2026, so the product had several months of closed testing before general use.

How a Make Agent Runs Inside a Scenario

Make agents live inside scenarios, the visual workflows Make users already build. The agent handles the steps that need judgment, and fixed modules handle the rest.

Adding an Agent and Its Tools

A builder places an agent on the canvas and writes its instructions. Then they add tools: any Make app or module. For each tool, the builder decides which inputs the AI fills in and which stay fixed. A support agent might get a Zendesk search tool where the AI writes the query, and a Slack tool where the channel is fixed.

The Reasoning Panel

While the agent runs, the Reasoning Panel shows how it is thinking, which tools it calls and why it chose each path. Make’s launch post says “nothing runs behind the scenes.” For an operations lead, this is the main difference from agents that return only a final answer.

Testing in the In-Canvas Chat

Builders can chat with the agent on the canvas to test it, question a decision or refine its instructions. Changes take effect in the same place the agent runs, so there is no separate test environment to keep in sync.

Files, Context and the Agent Library

Agents accept and produce files such as PDFs, images and CSVs. Users can upload files as context, so a policy document or price list can guide the agent without a retrieval workflow. A Library of Agents offers ready-made examples for inventory management, research, triage and reporting, and teams can share their own agents and full scenarios.

An Invoice Triage Agent, Step by Step

A finance team receives supplier invoices as PDFs by email. A Make scenario watches the inbox and passes each PDF to an agent. The agent reads the invoice, checks the supplier against a Google Sheet of approved vendors and decides where it goes: matched invoices go to the accounting app, unknown suppliers to a Slack channel for review. Fixed modules handle the email trigger and the final posting. The finance lead checks the Reasoning Panel when an invoice is routed in an unexpected way and adjusts the instructions in the chat.

Make Free, Core, Pro, Teams and Enterprise for Agents

AI Agents come with every paid Make plan. Make bills by credits, and the entry prices below come from Make’s comparison page.

Credit Budgets for Three Agent Workloads

Core’s entry price works out to $0.0009 per credit: $9 for 10,000. Most actions use one credit, so a five-step scenario that runs 1,000 times a month uses about 5,000 credits and fits inside Core.

Agent steps are harder to predict. Make says features that use its AI Provider may use more than one credit per step, without listing a rate. A team running an agent 2,000 times a month should expect agent steps to drive the bill and should test on the Free plan’s 1,000 credits first.

A larger operations team on Teams at $29 a month starts with 10,000 credits and can turn on auto-purchase, which buys 10,000 more before the balance runs out. Organizations with fewer than 20 employees may get 50% off through Make’s startup plan.

Credit Rules to Know

Error handler modules and the Router use no credits. A single scenario run can use anywhere from two credits to thousands, depending on what it does. Annual prepayment costs less than monthly billing.

Make AI Agents Against n8n, Zapier and Lindy

All four let non-developers or light coders build agents connected to business apps.

Make has the lowest paid entry price here. Its distinct feature is the Reasoning Panel on the same canvas as the workflow. n8n suits teams that need to self-host, Zapier suits teams already deep in its ecosystem and Lindy suits teams that want finished assistants. Lindy costs the most per seat but ships finished assistants for email and meetings, while Make and n8n ask the builder to design the flow. Teams comparing Make and Zapier should count the steps in their busiest workflow and price it on each, since Make bills by credits and Zapier by tasks. See our no-code agent builders for more.

Keeping Make Agents Inside Set Boundaries

An agent that can call any of thousands of apps needs limits. Make gives builders three. First, the builder picks which tools an agent can use; an agent with no email tool cannot send email. Second, for each tool the builder decides which inputs the AI may fill and which are fixed, such as a set Slack channel or a locked spreadsheet ID. Third, fixed modules can sit before and after the agent, so a human approval step or a filter can stop an action before it reaches a live system.

The Reasoning Panel then records why each call happened. For teams that must explain automated decisions to finance or compliance, that record is the main control. Enterprise customers can negotiate custom terms, and Teams plans add team features for sharing agents and scenarios.

A First Week With Make AI Agents

Day one: open a Library of Agents example close to your use case, such as triage or reporting, and run it on test data on the Free plan.

Days two and three: pick one existing scenario with a messy router or many filters. Replace that decision with an agent, give it two or three tools and fix every input that should not change.

Days four to seven: run the scenario on real inputs, read the Reasoning Panel for each run and note credit use per run. If the agent makes the right call often enough and the credit cost per run is acceptable, keep it and move to the next decision-heavy step.

Teams That Gain Most From Make AI Agents

Existing Make users gain the most. Their scenarios, connections and team habits carry over, and an agent can replace a tangle of routers and filters with one step that makes the call.

Marketing and research teams that pull information from many sources fit well too. An agent can look up a company in a CRM, read its website through a scraping module and write a summary to a sheet, choosing which source to check first based on what it finds.

Operations teams that need to explain automated decisions also fit. The Reasoning Panel gives a record of why an agent routed a ticket or flagged an invoice. Small teams on a tight budget can start on Free or Core and grow.

Builders Who May Look Elsewhere First

Developers who want agents in code, with tests and version control, will prefer a code-first framework. Teams that need to forecast AI costs to the dollar should note that Make does not list per-step credit rates for AI features.

Teams that need an agent to act across a desktop or inside sites with no Make app may need a browser agent, since Make agents call apps and modules through their connections.

Newcomers to automation should budget time for learning. G2 reviewers mention a learning curve and unclear error messages, and agents add another layer to learn on top of scenarios.

Automation Depth and Learning-Curve Notes From G2

Make holds 4.6 stars from 458 G2 reviews. Reviewers praise how much they can automate, the visual builder and how many apps connect without code.

The complaints are specific. Seven reviewers say error messages are hard to diagnose, and 13 mention the learning curve under two related themes. Five find module configuration complex, and four cite API documentation gaps and slow support. These reviews cover Make broadly. Make could add agent-specific case studies with numbers, such as hours saved per month, to help buyers judge the new agents.

Make Agent Milestones Since the Waitlist

• October 2025: Make opens a waitlist for its next-generation AI agent builder.
• February 11, 2026: Make releases next-generation AI Agents with the Reasoning Panel and Library of Agents.
• 2026: Make’s comparison page lists AI Agents on all paid plans and says Maia by Make is launching in early 2026.

Make AI Agents Score on Reach and Visibility

Make AI Agents score 9.1/10. Integrations lead, with more than 3,000 apps as tools. Pricing value is high, with a free plan and paid tiers from $9. Autonomy scores a little lower because agents work inside scenarios that people design, which is also what makes them predictable.

For teams already running Make, adding an agent to an existing scenario is the lowest-effort way to try agentic automation. Pick one decision-heavy step, replace it with an agent, read the Reasoning Panel for a week and compare credit use before and after.

What Automation Builders Ask About Make AI Agents

Are Make AI Agents Included in My Plan?

Make says AI Agents are included on all paid plans. The Free plan has 1,000 credits a month to try them.

How Many Credits Does an Agent Step Use?

Most Make actions use one credit. Make says features that use its AI Provider may use more, without listing a fixed rate, so test with real runs.

Can Make Agents Work With Files?

Yes. Agents accept and produce files such as PDFs, images and CSVs, so they can read an invoice, analyze a spreadsheet export or return a report file to the next step in the scenario.

Can I See Why the Agent Made a Decision?

Yes. The Reasoning Panel shows in real time how the agent reasons, which tools it calls and why it picked each path.

Which Apps Can a Make Agent Use?

Any Make app or module can be added as a tool. Make cites more than 3,000 apps.

What Is Maia by Make?

Make’s comparison page lists Maia by Make as launching in early 2026 alongside AI Agents and Make Grid. Buyers should check Make’s product pages for its current status and features.

Do I Need to Build a RAG Pipeline for Agent Context?

No. Make lets users upload files as agent context without building custom retrieval workflows.