agentsAgent Management

9 MCP tools.

Machine-readable: /reference/api/tools/agents.json

Tools

create_agent_task destructive

Create an agent task and add it to the queue for later processing. This is the "passive workflow" - task will be picked up when checking for pending work.

category
agents (Agent Management)
permissions
platform.super_admin

Parameters

agent_type : string
The agent type identifier
prompt : string
The prompt/instructions for this task
context : object (optional)
Additional context data for the task
priority : string (optional)
Task priority level — default: "normal" — enum: low, normal, high, urgent
assign_to_self : boolean (optional)
Assign the task to yourself. If False, task is unassigned. — default: true

disable_agent destructive

Disable an agent for your organization. Disabled agents cannot be executed until re-enabled.

category
agents (Agent Management)
permissions
agents.change_agentconfig

Parameters

agent_type : string
The agent type identifier to disable

enable_agent destructive

Enable an agent for your organization. Once enabled, the agent can be executed via run_agent or queued tasks.

category
agents (Agent Management)
permissions
agents.change_agentconfig

Parameters

agent_type : string
The agent type identifier to enable (e.g., "eob_matching", "expense_categorization")
settings : object (optional)
Optional settings for the agent (e.g., {"auto_process": true, "confidence_threshold": 0.85})

get_agent_history read-only

View execution history for agents. Can filter by agent, date range, status, and tier.

category
agents (Agent Management)
permissions
platform.super_admin

Parameters

agent_type : string (optional)
Filter by agent type. If not provided, shows all agents.
status : string (optional)
Filter by execution status — enum: pending, running, completed, failed, cancelled, waiting_approval
success_only : boolean (optional)
Only show successful executions — default: false
tier : string (optional)
Filter by execution tier — enum: tier1_gemini, tier2_claude, tier3_human
days_back : integer (optional)
Number of days to look back — default: 7
limit : integer (optional)
Maximum number of executions to return — default: 20

get_agent_status read-only

Get detailed status of a specific agent including configuration, recent executions, and queue depth.

category
agents (Agent Management)
permissions
platform.super_admin

Parameters

agent_type : string
The agent type identifier
include_recent_executions : boolean (optional)
Include recent execution history — default: true
execution_limit : integer (optional)
Number of recent executions to include — default: 5

get_pending_work read-only

Check if platform has pending work for Claude to process (agent work queue)

category
agents (Agent Work Queue)
permissions
platform.super_admin

list_agents read-only

List all available agents with their status, configuration, and capabilities. Shows which agents are enabled for your organization and their usage statistics.

category
agents (Agent Management)
permissions
platform.super_admin

Parameters

module : string (optional)
Filter by module (healthcare, finance, crm, legal, spare_parts, etc.)
enabled_only : boolean (optional)
Only show agents that are enabled for your organization — default: false
include_stats : boolean (optional)
Include execution statistics — default: true

mark_work_complete destructive

Mark agent task as completed and log execution results

category
agents (Agent Work Queue)
permissions
platform.super_admin

Parameters

task_id : string
UUID of the task to mark as complete
success : boolean
Whether task completed successfully
result_summary : object (optional)
Summary of results (items_processed, actions_taken, etc.)
error_message : string (optional)
Error message if task failed

run_agent destructive

Execute an agent directly with a custom prompt. Creates a task and immediately executes it. The agent must be enabled for your organization.

category
agents (Agent Management)
permissions
platform.super_admin

Parameters

agent_type : string
The agent type identifier (e.g., "eob_matching", "expense_categorization", "lead_scoring")
prompt : string (optional)
Custom prompt for this execution. If not provided, uses the agent default prompt.
context : object (optional)
Additional context data for the execution (e.g., {"claim_ids": [...], "date_range": "last_week"})
priority : string (optional)
Task priority level — default: "normal" — enum: low, normal, high, urgent