Use this tool for full lifecycle management of monday platform agents: create, read, update, delete, change state, and run. monday platform agents are user-built work orchestrators — each has a profile (name, role, avatar), a goal, and a markdown execution plan. Agents in the ACTIVE state can be triggered automatically. They are not local LangChain or MCP agents.

🚧
This tool runs against the monday.com dev (preview) API schema. The delete action is permanent and irreversible — when a user refers to an agent by name, always run action: "get" first to confirm the correct agent_id before deleting.

A single action parameter selects the operation. Only pass the fields that apply to the chosen action.

Action	Required fields	Description
`create`	`prompt`	Create an AI-generated agent. The platform generates the profile, goal, and plan from the prompt.
`create_blank`	—	Create a manually defined agent (pass `name`, `role`, etc.).
`get`	`agent_id` (omit to list owned agents)	Fetch one agent by ID, or list the agents you own.
`update`	`agent_id` + at least one mutable field	Modify `name`, `role`, `role_description`, `plan`, or `agent_model`.
`delete`	`agent_id`	Permanently delete an agent (irreversible).
`activate`	`agent_id`	Transition the agent to `ACTIVE`.
`deactivate`	`agent_id`	Transition the agent to `INACTIVE`.
`run`	`agent_id`	Manually enqueue an agent run (fire-and-forget).

📘
Created agents start in the INACTIVE state. Follow up with action: "activate" using the returned agent_id before the agent can be triggered. Agent state is one of ACTIVE, INACTIVE, ARCHIVED, or FAILED (DELETED only appears as the return value of delete).

Parameters

Parameter	Type	Required	Description
action	`enum`	Yes	One of `create`, `create_blank`, `get`, `update`, `delete`, `activate`, `deactivate`, `run`.
agent_id	`string`	Conditional	Required for `update`, `delete`, `activate`, `deactivate`, `run`. Use with `get` to fetch a specific agent; omit it with `get` to list owned agents.
prompt	`string`	Conditional	Required for `create`. Plain-language description of what the agent should do.
agent_model	`enum`	No	Used with `create` or `update`. Omit unless the user explicitly names a valid monday-supported model.
name	`string`	No	Used with `create_blank` or `update`. Display name of the agent.
role	`string`	No	Used with `create_blank` or `update`. Short role title (e.g. "Customer Success Bot").
role_description	`string`	No	Used with `create_blank` or `update`. Detailed description of the agent role.
avatar_url	`string`	No	Used with `create_blank`. HTTPS URL of the avatar. Prefer `dapulse-res.cloudinary.com` or `cdn.monday.com`.
gender	`enum`	No	Used with `create_blank`. `male` or `female` — a hint for the generated avatar/name when profile fields are omitted.
background_color	`string`	No	Used with `create_blank`. Lowercase hex, e.g. "#9450fd".
user_prompt	`string`	No	Used with `create_blank`. Stored as metadata; not used for AI generation.
plan	`string`	No	Used with `update`. New step-by-step execution plan in markdown.

Examples

Create an AI-generated agent:

{
  "action": "create",
  "prompt": "Run my daily standup every weekday at 9am."
}

Activate it after creation:

{
  "action": "activate",
  "agent_id": "7"
}

List the agents you own:

{
  "action": "get"
}

Related tools

agent_catalog — browse available trigger types and skills before wiring them to an agent.
manage_agent_triggers — manage which triggers fire this agent automatically.
manage_agent_skills — manage which skills this agent can perform.
manage_agent_knowledge — manage which boards and docs this agent can access.

Programmatic equivalent

This tool maps to the agent mutations and queries on the monday.com dev (preview) API schema (create_agent, create_blank_agent, agents, update_agent, delete_agent, activate_agent, deactivate_agent, run_agent). There is no stable public GraphQL equivalent yet.