monday.com · Solution guide
monday.com AI Agents for monday CRM: Built-In Agents, AI Credits and Custom Builds
Written scope and quote after the call. NDA on request. You own the code.
Quick answer
monday.com AI agents are monday's own agents (the AI Lead Agent, the AI Sales Agent and the agent builder) plus custom agents built on the monday API, webhooks and automations. They run on AI credits: about 10 to 250+ credits per agent run, 150 credits per five-minute AI Sales Agent call, with a one-time trial of 6,000 credits on most monday CRM plans. This page shows what each option does inside monday CRM, when the built-in agent is enough, when a custom build is worth it, what it costs to run and how we build one.
Builds like this
Part of the guide: CRM AI agents
Key takeaways
- monday.com's native agent options for monday CRM are the AI Lead Agent and the AI Sales Agent (both in beta), the AI agent builder for your own agents, and Bring Your Own Agent (beta) for agents hosted elsewhere (monday.com support, checked Oct 31, 2026).
- Agents run on AI credits. monday lists agent runs at about 10 to 50 credits for simple tasks up to 250+ for extra complex ones, and the CRM AI Sales Agent at 150 credits per five-minute call and 30 per follow-up SMS (checked Oct 31, 2026).
- monday CRM's plan article gives Basic, Standard and Pro accounts a one-time trial of 6,000 AI credits and Ultimate 12,000; after that you buy an AI credits add-on (checked Oct 31, 2026).
- A custom agent is worth building when the job spans systems monday's agent tools don't reach, needs your own model or logic, or needs approval steps and logs you control. External agents connect through Bring Your Own Agent or a signed webhook, and act through the GraphQL API.
- Every API or Platform MCP call a custom agent makes counts toward the account's daily API limit: 1,000 calls on Basic and Standard, 10,000 on Pro, 25,000 on Enterprise (monday developer docs, updated Oct 28, 2026).
monday.com AI agents are AI workers that act on your monday CRM boards: they read leads, deals and contacts, decide what to do next within rules you set, then create or update items, assign owners, draft messages or place calls. monday ships its own (the AI Lead Agent, the AI Sales Agent and a general agent builder), and teams with jobs those can’t reach build custom agents on the monday API, webhooks and automations. A typical first build is one agent with one or two jobs on the Leads or Deals board, tested on real records before it is allowed to write. This page covers both routes, what they cost to run in AI credits and API calls, and how our in-house engineers build and guard a custom one.
Sources: monday.com support articles on monday CRM plan types, AI credits and the CRM AI Sales Agent, and monday’s developer rate-limits page, all checked Oct 31, 2026. Rates are monday’s published guidance and can change.
What are monday.com AI agents, and which ones exist today?
monday’s AI Agents overview (checked Oct 31, 2026) describes agents as built inside monday.com and acting “within the context of your work”: they monitor activity, make decisions from your rules and priorities, and execute tasks on boards. They differ from automations, which follow fixed if-then rules, and from monday sidekick, which is a chat assistant you ask for help on demand. There are five things people mean when they search for monday AI agents, and they are not interchangeable.
| Option | What it is | Where it runs | Status (checked Oct 31, 2026) |
|---|---|---|---|
| AI Lead Agent | Expert Agent that sources, adds and enriches leads matching your ideal customer profile, on a recurring schedule | Leads board in monday CRM | Beta |
| AI Sales Agent | Expert Agent that places outbound voice calls and SMS follow-ups, logging summary, transcript and outcome on the item | Any board with a phone column, triggered by an automation | Beta, US only for calling |
| AI agent builder | Build your own agent: instructions, board and file access, tools, skills, jobs with triggers, AI model, channels such as WhatsApp and Telegram | monday.com, AI Agents tab | Live on the monday AI platform; monday says availability on all products is “coming soon” |
| Bring Your Own Agent | Connect an agent hosted on another platform so it can be chatted with, mentioned and assigned items | Your platform, governed in monday.com | Beta |
| Agent Factory | Standalone agent product outside monday.com | Separate product | monday says it cannot connect to a monday.com account at this time |
Sources: monday.com support, AI Agents on monday.com, AI Lead Agent, AI Sales Agent, Bring your external agent into monday.com and Agent Factory, all checked Oct 31, 2026.
The monday CRM pricing page (checked Oct 31, 2026) also lists an outreach caller agent, a lead sourcer agent, a meeting prep agent, build-your-own agents, the AI Notetaker, sidekick, the vibe app builder and AI-powered data fields on Standard, Pro and Ultimate, with the AI workflow builder on Pro and Ultimate. monday’s own FAQ says admin permissions are needed to create or manage agents. If the AI Agents tab isn’t in your CRM account, the “coming soon” note above is the likely reason, so check before you plan a build around the builder.
monday’s marketing page, monday CRM AI agents (checked Oct 31, 2026), uses different names and lists its pre-built agents for Pro and above. Lead Sourcing there is the AI Lead Agent in the support articles, Calling is the AI Sales Agent, and Meeting Prep is the meeting prep agent on the pricing page. We found no separate support article for Pipeline Monitor, Campaign Intelligence or Campaign Builder, so treat those as marketing names for agents configured on the same platform, and confirm they appear in your account before you plan around them. Note the plan wording differs too: the AI agents page says Pro and above, the pricing page lists agents on Standard, Pro and Ultimate.
The jobs an AI agent does inside monday CRM
It reads items on boards it has been given, applies your instructions, and writes back: new items, column values, owners, statuses, updates and drafted messages. The useful way to scope one is by job: a trigger, the records it reads, the records it writes, and what a human checks. These are the jobs we see asked for most on monday CRM.
| Job | Trigger | Reads | Writes | Native or custom? |
|---|---|---|---|---|
| Lead sourcing to your ICP | Schedule | ICP settings, Leads board | New leads in a “leads to review” group, mapped columns | Native AI Lead Agent covers it |
| Speed-to-lead call | Status changes to a value you pick | Phone column, lead fields | Call summary, transcript, outcome on the item timeline | Native AI Sales Agent, US numbers only |
| Lead triage and routing | New item on Leads | Form answers, source, territory columns | Owner, priority, status, a drafted first reply | Agent builder, or custom if routing rules live in another system |
| Meeting prep brief | Meeting on the calendar | Contact, account, deal history, emails | A brief on the item or in an update | Native meeting prep agent where your plan includes it |
| Stale deal review | Weekly schedule | Deals board stage, last activity date, amount | Update with next step, flag column, owner notification | Agent builder for one board; custom for cross-system checks |
| Won-deal handoff | Deal status to Won | Deal, account, contact, line items | Onboarding item, owner, checklist, a note to delivery | Automation plus agent, or custom when billing or ERP must be updated |
| Data hygiene | Schedule or new item | Contacts, Accounts | Merged or flagged duplicates, filled gaps | Custom when enrichment comes from your own sources |
Source: aibrevo scoping notes for monday CRM agent requests, Oct 31, 2026; native coverage from the monday.com support articles cited above. For the cross-board handoffs that agents are often asked to fix, the monday.com automation recipes page explains where plain automations already do the job and where they create duplicates.
When is monday’s built-in agent enough, and when do you need a custom one?
Start with the built-in agent when the job stays inside monday, the data it needs is on boards you can grant, and the tools it needs are on monday’s list. The agent builder already connects to Gmail, Outlook, Slack, Microsoft Teams, Google and Outlook calendars, HubSpot, Salesforce, Google Drive, Dropbox, SharePoint, Calendly, Twilio, Xero, GitHub and others, according to monday’s build and configure article (checked Oct 31, 2026). It also separates read tools, which are active once connected, from write tools, which are added inactive until you switch them on.
| Situation | Built-in agent | Custom agent |
|---|---|---|
| One board, one clear job, tools on monday’s list | Usually enough | Not needed |
| Outbound calls to US leads from a board | AI Sales Agent | Only if you need non-US calling or your own telephony |
| Data lives in a system the tool list doesn’t cover (ERP, billing, a product database) | Can’t reach it | Needed |
| You need a specific model, your own prompts in version control, or your own evaluation tests | Model choice exists, but configuration lives in monday | Needed |
| Each write must wait for a named person’s approval | Partly, through write-tool activation and guardrails | Built as an approval column plus a gated mutation |
| You need logs in your own system for audit | Chats and activity pane inside monday | Your own logs, plus monday’s |
| The agent must act on hundreds of items an hour | Credits per run add up | API cost and limits need planning either way |
Source: monday.com support and developer docs, checked Oct 31, 2026; the decision column is aibrevo’s guidance. A custom agent is not automatically better. It costs engineering time to build and keep running, and monday’s own agents get updates you don’t pay for. We recommend custom only for the rows on the right that apply to you.
Three ways to build a custom monday.com AI agent
There are three patterns, and the choice depends on where the agent’s logic should live.
- Configured inside the agent builder. We write the instructions, grant only the boards it needs, connect tools, split the work into jobs with their own triggers, and set guardrails such as “only email contacts on this board.” The logic lives in monday. This is the fastest route and still benefits from proper scoping and testing.
- An external agent connected to monday. The agent runs on our infrastructure or yours and joins monday as its own identity. monday’s external agent guide (checked Oct 31, 2026) describes the flow: monday sends a signed webhook when the agent is chatted with, mentioned or assigned an item; the service verifies it with a signing secret; and the agent acts through the GraphQL API with its own API token. monday waits about 30 seconds for a chat reply, so slower work is acknowledged first and posted back as an update. The guide says these operations currently need the
devAPI version, and the agents query is in version 2027-01, a release candidate until January 15, 2027, so we pin versions and plan for changes. - A service driven by monday automations or webhooks. A board event calls our service, which uses a model to decide and then writes back through the API. This is the right shape when the “agent” is really one judgment step inside a bigger integration, such as classifying inbound leads before a sync to another system. The API side is covered in our monday CRM API integration guide, and wider sync work is on our CRM integration services page.
monday also hosts a Platform MCP server (checked Oct 31, 2026) that AI clients such as Claude, ChatGPT and Copilot can use to read and write board data. It is useful for internal assistants. Its tool calls run as GraphQL requests and count toward the daily API limit, which matters once an agent runs on a schedule.
How we build and guard a monday CRM agent
aibrevo’s in-house engineering team of 16, led by co-founder and CTO Shivam, builds custom CRM apps, integrations and AI agents on the four CRMs we implement: GoHighLevel, HubSpot, monday.com and Pipedrive. Projects are scoped and quoted per build after the call, and the written scope names the agent’s jobs, the boards and permissions it gets, the guardrails and what we hand over. We publish no price list for agent builds. The board structure the agent relies on is often the first thing to fix, which is part of our monday.com implementation work. The team follows the same order on every agent build.
1. Discovery
One job written as trigger, inputs, decision, outputs and who checks the result. You can check: a one-page job spec you agree with.
2. Data model and permissions
Which boards and columns the agent can read and which it can write. monday says shared agents run with the owner's context and permissions and that ownership cannot be transferred, so we build under a dedicated account, not a person who may leave. You can check: a permissions table.
3. Build
Instructions, tools and jobs in the builder, or the external service, webhook handler and API calls for a custom agent. You can check: a demo on a copy of your board.
4. Test on real records
Runs against a sample of real leads or deals in a duplicate board, with expected outputs written down first. You can check: a test log of inputs, outputs and misses.
5. Guardrails
Write tools off until approved, an approval status column for anything customer-facing, a cap on runs per day, logging of every action, and backoff on the API's retry_in_seconds. You can check: the guardrail list and the log.
6. Go live and review
Turn on for one team, read the first two weeks of runs, adjust. You can check: a short review with credit and API usage.
Personal data needs a decision before the build, not after. A built-in agent processes data under monday’s terms and the model you pick in the builder. An external agent sends board data to wherever it is hosted and to its model provider, so we list which fields leave monday and keep the rest out of the prompt. If you handle health data, read our note on whether monday.com is HIPAA compliant first.
What does a monday AI agent cost to run?
Two meters run at once: AI credits for monday’s own AI, and API calls for anything custom. Seats come on top.
| Item | Rate | Source, checked Oct 31, 2026 |
|---|---|---|
| monday agents, per run | Simple ~10 to 50, intermediate ~50 to 150, complex ~150 to 250, extra complex ~250+ credits | monday support, AI Credits |
| monday sidekick, per message | ~10 to 30 up to ~150+ credits | monday support, AI Credits |
| AI Notetaker | 120 credits per meeting hour | monday support, AI Credits |
| AI blocks | 8 credits per action | monday support, AI Credits |
| CRM AI Sales Agent | 150 credits per five-minute call, 30 per follow-up SMS, ~600 for a 20-minute call | monday support, AI Sales Agent |
| CRM trial credits | 6,000 one-time on Basic, Standard, Pro; 12,000 on Ultimate | monday support, CRM plan types |
| CRM seats | $12, $17, $28 per seat per month billed yearly (Basic, Standard, Pro) | monday CRM pricing page |
| Daily API calls | 1,000 Basic and Standard, 10,000 Pro, 25,000 Enterprise; API add-on available | monday developer docs, rate limits |
One caveat on the credit rates: monday’s AI Credits article says its billing model covers the monday AI work platform and “does not affect monday CRM.” The CRM side is governed by the trial credits and add-on described in the CRM plan types article, and the CRM pricing page says credits are included on Standard and above without stating a number. Treat the per-feature rates as monday’s best published guide for CRM and confirm against the Credits Usage tab in AI governance. Our monday AI credits explainer covers alerts at 80% and 100% and per-user limits.
Illustrative arithmetic: an AI Sales Agent that makes 40 five-minute calls uses 6,000 credits, the whole trial allowance on a Pro account. An agent builder job that runs at an intermediate ~100 credits per run, 20 times a working day, uses about 2,000 credits a day. Neither figure includes model choice, which monday says changes credit cost for agents. For a custom agent, monday’s AI credits may not apply at all, but your model provider’s usage fees and hosting do, and so does the daily API limit: an agent that reads and writes 10 times per lead on 100 leads a day needs 1,000 calls, the full allowance on Standard. Our daily API call limit page explains what counts and how to request more.
How long does a monday AI agent build take?
These are the ranges we use when scoping, and the written scope sets the real date: configuring and testing one built-in agent with one job usually takes days to about a week; a custom external agent with one job, its webhook handler, logging and approval step usually takes two to four weeks; and an agent with several jobs that also syncs another system can run four to eight weeks. The calendar depends most on how quickly someone on your side can review test runs. We give a firm timeline in the written scope after the call.
Checks before you switch an agent on
- Beta status. The AI Lead Agent, AI Sales Agent and Bring Your Own Agent are labelled beta in monday’s articles. Beta features can change behaviour and pricing.
- Calling rules. The AI Sales Agent calls only from the US and needs country codes on every number. Whether you may call a lead at all is a consent and telemarketing question for your counsel, not a monday setting.
- Board changes break agents. monday says the AI Lead Agent pauses on errors such as missing boards, groups, columns or permissions. Renaming a column can stop a job, so freeze the structure the agent depends on.
- Restricted columns. An agent can edit a restricted column only if it owns the board or is added to the column’s permissions, per monday’s FAQ. Decide this deliberately.
- Credit burn. Set per-user limits and watch the first week. A loop that re-triggers an agent on its own updates is the usual cause of a surprise bill.
- API versions. Custom agent operations sit on new API versions. Pin a version and read monday’s release notes before each quarterly change.
Who is a monday AI agent for, and who should wait?
It suits teams already running monday CRM with clean Leads and Deals boards, a repeated job that eats hours each week, and someone who will read the agent’s output for the first few weeks. It is a poor fit when the boards are still changing shape, when nobody owns the process the agent would run, or when one automation would do. If you are still choosing a CRM, compare first: HubSpot vs monday and monday vs Pipedrive cover the platform differences, and the wider CRM AI agents page compares agent options across the four CRMs we work on. If your agent needs its own screen inside monday, such as a review queue, that is monday.com app development.
Want an agent working on your monday CRM boards? Book a free 30-minute call with the job you want automated, the boards involved and the systems outside monday it touches, or send a message. We tell you whether monday's built-in agent covers it before quoting a custom build.
Sources
All pages opened on Oct 31, 2026.
- monday.com support, AI Agents on monday.com: agent types, builder tabs, governance, ownership, admin requirement, availability note.
- monday.com support, Build and configure your AI agent: knowledge and access, read and write tools, guardrails, supported integrations.
- monday.com support, monday CRM AI Lead Agent and monday CRM AI Sales Agent: beta status, setup, limits, credit use, US-only calling.
- monday.com support, AI Credits and AI Models and Credits: consumption rates, model choice, limit alerts, scope note.
- monday.com support, Available plan types on monday CRM: 6,000 and 12,000 trial credits, add-on.
- monday.com support, Bring your external agent into monday.com and Agent Factory: supported providers, standalone status.
- monday.com, monday CRM pricing: seat prices, AI features by plan.
- monday.com developers, Build an external agent, Agents reference, Platform MCP and Rate limits: webhook flow, API versions, MCP limit note, daily and per-minute limits.
