SoftTech AI Registry
monday.com · monday AI work platform

AI Agents

monday.com AI agents are configurable agents built inside monday.com through the AI agent builder. Within the boards, permissions and guardrails their creator configures, they monitor activity, evaluate context, and execute work end to end: creating and updating items, assigning owners, changing statuses, drafting messages, logging outcomes and executing follow-ups. Agents run from assignment and mention triggers, configured Jobs, schedules and chat requests, and can be extended with read and write tools that reach external systems.

Recorded characteristics

Function
monday.com documents AI agents as agents built inside monday.com that "act within the context of your work": they monitor activity, make decisions based on rules and priorities the creator defines, and execute tasks end to end on boards and workflows. Documented functions include understanding context from boards, data, docs, workflows and permissions; applying defined rules, priorities, tone and thresholds to triage, route, escalate and prioritise work; creating and updating items, assigning owners, changing statuses, drafting messages, logging outcomes and executing follow-ups; and running when specific events occur, at defined times, or on an ongoing cadence. Agents are configured in an agent builder with four tabs: Brain (instructions, knowledge and access, tools, skills, AI model), Jobs (separate tasks each with a trigger and scoped instructions, able to run in parallel or on a daily, weekly or monthly schedule), Channels (communication with people in external apps such as WhatsApp and Telegram) and Activity (run monitoring). Two default Jobs, When Assigned and When Mentioned, exist on every agent, and a user can also ask an agent for a one-off task. Agents retain memory across runs, produce stored artifacts, can work inside monday docs, and can join live meetings alongside AI Notetaker. This record covers the configurable AI-agent capability, including agents built with the AI agent builder. It is not a record of any individual agent: monday states that what a specific agent can do depends on the instructions, access, tools, skills and Jobs configured for it, so the documented example actions are the range of the capability rather than the behaviour of every agent. Expert Agents (such as the CRM AI Lead Agent and CRM Sales Agent), Org agents as a directory, monday sidekick, AI workflows, Agent Factory and externally built agents connected through Bring Your Own Agent are adjacent and outside this record.
Data access
Documented potential access, all of it configuration-dependent. In the Brain tab the creator grants the agent access to monday boards and uploads files (pdf, doc, txt, excel, csv) holding additional context such as FAQs or governance and legal information, and specifies for each grant whether the agent may only read board data or may also edit and add to it. monday states that agents "only access what you are allowed to access" and that the grant can be narrowed to the minimum needed, including granting the same viewing permissions as a particular member. Agents can use monday docs as a knowledge base and can write to a doc only when the doc's permissions are set to read-write. Optional access includes web search (a toggle), the AI Notetaker (meeting records, summaries, transcripts and action items), platform integrations and custom MCP endpoints, and external context through connected external files. monday notes that some tools connect via MCP and that their individual capabilities are not displayed in the interface. Nothing in the current documentation establishes automatic account-wide or workspace-wide access: every source above is one the creator connects.
Actions
Can take actions
External actions
Conditional
Human confirmation
Conditional
Permission basis
Mixed
Administrative control
Documented controls. Account level: an admin enables or disables monday AI for the account under Administration > AI Governance > Settings; the AI permissions tab (Enterprise plan only) controls which account roles may use each agent type (user agents, monday agents, third party agents, org agents, external AI connectors) and each AI feature, with per-agent settings or inheritance from the parent level; credit usage and per-capability and per-user usage limits are also managed there. monday notes that changing permissions does not stop agents already running. The AI Agents FAQ states that admin permissions are required to create or manage AI agents. Agent level: the creator sets instructions, knowledge and access with read or edit scope, tools, skills, AI model, Jobs and triggers, and channels; creators control where an agent may act (specific boards, spaces, teams) and what it may do within that scope. Write tools are added inactive and must be explicitly activated after review, and guardrails restrict a tool, for example to a specific Slack channel or to certain email recipients. Every configuration action taken through the agent chat offers an Undo, which covers configuration only and not content the agent has already produced. Agents can be paused at any time and continue to be editable while paused. Owners assign editors and viewers; ownership cannot currently be transferred. Sharing is per member, per team or organisation-wide through the Org agents directory, with shared members able to trigger, chat with and view settings but not edit configuration; a Review private content step discloses private boards and docs before sharing, personal tool connections require an explicit toggle before shared members may use them, and admins control which roles may share agents. Monitoring: the Activity tab logs every run with status, apps used and credits consumed, failed runs raise a notification with a What Went Wrong panel, board activity logs allow an agent's actions to be reviewed and undone, Live agent mode shows the agent working in real time, and agent memory can be managed, switched off entirely, or edited by asking the agent to remember or forget.
Default state
Not established
Availability
Every current AI agent article carries the availability banner "Available on monday AI platform, availability on all products coming soon", so as of September 2026 monday documents the capability as available on the monday AI platform with wider product availability still being rolled out rather than generally available across all monday products. Account admins must have monday AI enabled, and on Enterprise the AI permissions tab governs which roles may use each agent type. Shared agents are available to all account members except viewers and guests. Agents are also usable on mobile through the AI agents app. Tools on Demand, which builds a custom integration from a chat request, is documented as Beta.
Licensing
monday documents agent runs as consuming AI credits from the account's shared credit pool, with consumption depending on the complexity and depth of each task and attributed to the agent owner even when other members trigger the agent. Admins can review credit usage and set account, per-capability and per-user limits, and can buy more credits. No plan price for agents is documented beyond the Enterprise restriction on the AI permissions tab; monday does not publish an agent-specific licence in this documentation.
External model or provider
Not established as a specific model. monday documents a Model setting in the Brain tab where the creator chooses the agent's primary AI model and notes that additional models may be used by the agent to optimise performance, but the current articles do not name the models or the providers behind them.
Limitations and uncertainty
ESTABLISHED: agents make persistent changes inside monday (create and update items, assign owners, change statuses, draft messages, log outcomes, execute follow-ups); they run from assignment and mention triggers, configured Job triggers, recurring daily, weekly or monthly schedules and direct chat requests, including while the user is away; tools are classified Read or Write, with write tools covering send, create, edit and delete and being added inactive until the configurer explicitly activates them; paused agents cannot run but remain editable; the Activity tab records every run. CONFIGURATION-DEPENDENT: which actions any individual agent can take, which boards and files it can read or edit, which external tools or MCP endpoints it holds, which Jobs and schedules exist, and whether behavioural rules such as "always ask before sending an email" are stored in its memory. NOT ESTABLISHED: the operational default of the capability itself — monday documents that a newly created agent is active as soon as it is created with no manual activation step, but does not document whether AI Agents are on or off by default for an account or workspace, and the rollout banner and admin AI permissions prevent that being inferred; a single account-wide runtime permission rule — the current documentation describes the creator's own access as the ceiling, per-agent read or edit scopes set at configuration time, the agent being addressable as a grantee in board and column permissions, shared members inheriting the agent's context, and admin role controls over agent types, which together make the runtime basis mixed rather than any single one of those; the underlying models and providers; and any account-level approval workflow that applies to consequential actions independently of what the creator configures. Boundaries: access to external context through connected files or the web does not by itself establish external action, which rests on activated write tools and channels; "operate within governance controls" and "within your permissions" describe boundaries set at configuration time rather than a per-run identity; and the documented example actions describe what the capability supports, not what every agent does.

Evidence