ReferenceMCP serverMCP tools

Updated 27 Oct 2026View as Markdown

MCP tools

Every tool on the mindset-admin MCP server, grouped by job, and the two tools on the users' mindset server.

Org admins reach the mindset-admin server from a coding agent or AI client. Members reach the mindset server, which carries two tools. Setting either up is on Start with your coding agent and Use your agents from AI clients.

Mindset checks the caller's permissions on every call. Environment-scoped tools take an optional envSlug that points that one call at another environment. Org-level tools refuse it.

Guidance#

Tool	What it does
`get_information`	Returns the topic index, or one topic's body: what the platform is, the authoring skill for each resource type, the DSL reference and capability boundaries
`find_examples`	Finds curated authoring examples close to what you describe
`platform_issue`	Reports something on the platform you couldn't make work

Agents#

Tool	What it does
`create_agent`	Creates an agent. It starts as a draft and unavailable
`get_agent`	Reads one agent by its @handle
`list_agents`	Lists agents, filtered by whether they're available or by a search
`update_agent`	Changes an agent's prompt, model, script, availability or schedules. A change to its configuration mints a version, staged or activated
`archive_agent`	Takes an agent out of use and keeps it listed
`delete_agent`	Removes an agent from every list. An admin can't undo it
`agent_config_admin`	Sets what an agent may use (operations, connections, functions, widgets and other agents), and publishes, activates or rolls back that membership as versions
`agent_config_versions`	Lists, reads, compares and activates versions of what an agent says, and agrees a version's writes before it's active
`list_agent_prompt_history`	Reads every recorded wording of each prompt field. Read only

Resources#

Scripts, functions, widgets, connections, operations, triggers and test parameter sets all ride the same verbs, chosen by type.

Tool	What it does
`create_resource`	Creates a resource of the given type
`get_resource`	Reads one resource
`list_resources`	Lists resources of a type
`edit_resource`	Changes a resource, including an operation's name, description and who may call it
`publish_resource`	Publishes a resource so agents can use it
`unpublish_resource`	Takes a published resource out of use
`delete_resource`	Deletes a resource
`list_resource_versions`	Lists a resource's versions
`activate_resource_version`	Makes an earlier version of a function or widget active. A script's version is pinned by its agent's version instead
`render_resource`	Describes in English what a function or widget document does, including what happens when a step fails. Saves nothing
`render_function_doc`	The same reading, for a function
`render_script_doc`	Describes every phase, gate and route of a script
`render_widget_doc`	Describes a widget's components, bindings and rules

Connections#

Tool	What it does
`probe_endpoint`	Asks a URL what it is (an MCP server or a plain HTTP host) and what sign-in it wants, before a connection is created
`introspect_api`	Looks for a machine-readable API description on a connection's own host and proposes operations. Enables nothing
`request_capture`	Starts a credential capture and returns a reference and a capture URL for a person to open. No tool takes a secret value
`authorize_connection`	Returns an OAuth sign-in link for a person to open
`bind_connection_credential`	Binds your org's own credential to a connection an operator published
`revoke_secret`	Stops a credential working at once, for every connection and agent that holds it
`govern_tool`	Enables, disables, hides or restricts one operation to admins, and sets whether it reads or writes. A write that needs approval stays waiting

Testing#

Tool	What it does
`test_resource`	Runs a function, script, widget, connection operation or trigger and reports what happened. A run that writes is refused until an admin gives the go-ahead for its write list
`get_run`	Reads one run's trace, by run ID or conversation ID
`agent_testing`	Writes and runs behavior tests, compares a model or prompt change, and reads the verdict. Behavior tests refuse every write

Runs and conversations#

Tool	What it does
`call_agent`	Hands an agent a task and returns its result. An admin can call an agent that isn't available, or name a version to try
`list_callable_agents`	Lists the agents you can call, each with its description and the operations it holds
`list_dashboard_agents`	Lists the agents published to people in the Hub
`start_conversation`	Starts a Hub conversation with one agent
`list_my_conversations`	Lists your own Hub conversations
`conversation_admin`	Resumes, renames or deletes a conversation
`conversation_fetch`	Reads what people and agents did: conversations, messages, a run's events and the model's verbatim input
`list_my_builder_conversations`	Lists your past conversations with one of the Studio's builder agents
`get_my_builder_conversation`	Reads one of those conversations in full
`run_admin`	Previews or stops runs in flight for one run, agent or connection. Changes no configuration
`org_state`	Reports what's unresolved in the org now, such as who still has a credential to supply or what is broken

Triggers and schedules#

Creating, changing and removing schedules and webhook bindings goes through update_agent, and takes effect when the version is activated.

Tool	What it does
`list_triggers`	Lists one agent's schedules and webhook bindings, or every one in the environment
`get_trigger`	Reads one trigger
`list_trigger_runs`	Lists a trigger's runs
`list_trigger_brief_history`	Reads every recorded wording of a trigger's standing instruction. Read only
`list_schedules`	Lists a function's old schedules. Function schedules no longer fire: an agent's schedule is a trigger
`get_schedule`	Reads one of those old function schedules
`list_schedule_runs`	Lists an old function schedule's past firings
`list_webhook_deliveries`	Lists what a webhook source received, to find out why nothing arrives

Costs#

Tool	What it does
`costs`	Reads what the org spent on model calls, the same figures as Govern → Monitor → Cost

Audit#

Tool	What it does
`query_audit`	Queries the audit trail
`get_audit_event`	Reads one audit record
`render_chart`	Draws a bar, line or pie chart of the audit trail. Mindset counts the events itself
`audit`	Reads data residency and retention, the compliance controls, personal data detections and erasures, and verifies the tamper-evidence chain. Its one write records an attestation

The org#

Tool	What it does
`member_admin`	Lists, invites, enables, disables and removes members, and changes their role
`org_admin`	Sets audit and conversation retention, reply language, personal agents and personal data detection, lists and creates environments, and revokes an embedded user's session
`org_domain_admin`	Sets email-domain auto-grouping, directory listing and self signup, and the websites allowed to sign visitors in to an embedded agent
`get_pii_policy`	Reads the org's personal data policy
`set_pii_policy`	Sets the org's personal data policy
`update_env`	Renames an environment. Its slug never changes
`fleet`	Creates, suspends, restores and purges orgs on the instance. Platform operators only
`platform_org_members`	Looks after the people in a customer org. Platform operators only

The users' `mindset` server#

Tool	What it does
`list_callable_agents`	Lists the agents available to you, each with its description
`call_agent`	Hands one of them a task and returns its result