To build a monday.com AI agent, open the AI Agents tab in the left pane, click + New agent, describe the job in plain words, then set which boards it can read or edit, which tools it may use, and when it runs. Agents run on AI credits. Paid plans bought on or after May 6, 2026 include a monthly credit minimum: 1,000 on Basic, 2,000 on Standard and 3,000 on Pro.
I set up monday.com for operations and project teams. Agents are useful, but they cost credits every time they work, and a vague agent burns more. This guide follows monday.com’s agent builder article and its credits article, checked on October 15, 2026.
First, which “agent” do you mean?
monday.com uses the word in two places. Mixing them up is the most common mistake I see.
- monday agents live inside monday.com. They work on your boards and workflows. This guide is about these.
- Agent Factory is a separate product. monday.com’s Agent Factory article says it does not need a monday account, has its own terms and pricing, and cannot connect to your monday.com account at this time. It is built for things like phone agents.
If you want an agent to update your boards, use monday agents.
What it costs
monday.com now sells seats plus AI credits. For accounts that joined on or after May 6, 2026, credits are part of the purchase. The pricing model article lists the minimums:
| Plan | Seat price (yearly) | Minimum AI credits per month |
|---|---|---|
| Free | $0 (up to 2 seats) | No AI features |
| Basic | $9 | 1,000 |
| Standard | $12 | 2,000 |
| Pro | $19 | 3,000 |
| Enterprise | Custom | Custom |
On the pricing page, 1,000 credits are shown as $10. When I checked, a special offer took that amount back off the monthly total. Offers change, so check yours.
How many credits does an agent use? monday.com gives ranges per task:
- Simple (status updates, quick checks): about 10 to 50 credits.
- Intermediate (multi-step work, summaries): about 50 to 150.
- Complex (research, screening, reports): about 150 to 250.
- Extra complex (batch jobs, deep web research): 250+.
One run can include several tasks. monday.com’s own example, a weekly “Sprint Reviewer,” uses about 63 credits per run. A request router running 15 times a day uses about 5,400 credits a month. That is far more than the Pro minimum, so budget before you switch it on. The same credit pool also pays for monday vibe apps, covered in our vibe widgets guide.
Step 1: Create the agent
- Click AI Agents in the left pane.
- Click + New agent.
- Describe the agent in plain text, start from a template, or pick + Start from blank. Templates cover Productivity, Marketing, Project management, Operations, Sales, HR, and Product and Engineering.
- Give it a name, job title and avatar. Click Onboard.
Note: once onboarded, the agent is active right away with the instructions you gave. If you want to build quietly first, use Pause agent from the three-dot menu. A paused agent can be edited but cannot run.
Step 2: Tighten the instructions
The agent chat helps you build. It suggests missing details and offers to run a test. After each change, you get an Undo option. Undo covers settings only, not content the agent already wrote.
Then open the Profile tab and read the Identity instructions. Make them specific. Here is a pattern that works:
“When an item on the Client Requests board is set to New, read the description. Set Priority and Type. Assign it to the right team lead using the Owners board. Post a one-line update on the item. Do not change any other column.”
Step 3: Set knowledge and access
In Knowledge and access, pick the boards the agent can use. For each, choose read only or read and edit. Grant the least you can. You can also upload files like FAQs or policies (PDF, doc, txt, Excel or CSV).
monday.com says it plainly: an agent cannot act outside what it is explicitly permitted to do. That is your main safety control.
Step 4: Add tools, with guardrails
Tools connect the agent to other apps. Ask for one in the chat, like “post to our #support Slack channel.” The agent finds the tool and walks you through the connection.
- Read tools (search, pull data) are active right after you connect them.
- Write tools (send, create, edit, delete) start inactive. You review and switch them on.
- Guardrails set limits, like one Slack channel only.
Built-in tools include web search, web browsing, image generation, AI Notetaker, and custom MCP connections. Turn web search off unless you need it. monday.com says it raises credit use.
Step 5: Set up Jobs
Jobs split the agent’s work into focused tasks, each with its own trigger and instructions. Every agent has two default Jobs: When Assigned and When Mentioned. You can add scheduled Jobs that run daily, weekly or monthly.
Example: a project agent with one Job that runs Friday morning to write a status update, and another that runs when an item turns Stuck.
Step 6: Test, then watch the Activity tab
Run a simulation before you trust it. After real runs, open Activity. monday.com says credit use per run shows there within about 15 minutes. Read what it did, then tighten the instructions.
Control the spend
Admins manage AI under Administration, AI Governance, per monday.com’s AI governance article:
- Credit usage: see use by feature and by user.
- Usage limits: set limits for the account, per AI feature, and per user. Each feature limit is marked hard or soft.
- Agent directory: see every agent, its owner, its board access and model. Pause or transfer agents here.
- Alerts fire at 80% and 100% of your credit limit.
- AI permissions by role are on Enterprise only.
Where AI blocks fit
Not every AI job needs an agent. AI blocks do one thing on an item, like summarize, categorize or extract. They cost 8 credits per action, and all actions on the same item within 24 hours count once. Subitems count separately. For simple sorting, a block inside an automation is cheaper and more predictable than an agent.
Where the Autopilot hub fits
For regular automations, the Autopilot hub shows account-wide health. Open More in the left pane and pick Automation hub. Its tabs are Health, Usage, Workflows and Connections. The Health tab gives an AI-generated root cause for failed automations. If rules keep failing, our monday.com automations fix guide can help.
Honest pros and cons
Pros: real board access with read or edit control, tools that start inactive for writes, clear Jobs, and good admin limits.
Cons: credits are hard to predict until you run for a few weeks. Role-based AI permissions need Enterprise. Agent Factory’s name causes confusion, and it does not connect to monday.com. For a side-by-side with rivals, see our AI features comparison. Plan costs are in our monday.com pricing guide.
Common questions
Is Agent Factory the same as monday AI agents?
No. Agent Factory is a separate product with its own terms and pricing. It cannot connect to your monday.com account at this time. monday agents work inside monday.com on your boards.
How many AI credits come with monday.com plans?
For accounts that joined on or after May 6, 2026, the monthly minimums are 1,000 on Basic, 2,000 on Standard and 3,000 on Pro. Enterprise is custom.
Are AI agents available on the free plan?
No. monday.com says AI features are for paying accounts only.
How do I limit what a monday agent can change?
Grant read-only access to boards it should not edit, keep write tools inactive until reviewed, and add guardrails like one Slack channel only.
How much does a monday AI block cost?
8 credits per action. All actions on the same item within 24 hours count once, and subitems count separately.
Checked against support.monday.com and monday.com/pricing on October 15, 2026. AI pricing changes often; confirm before you plan a budget.
Sponsored
Related guides
