Work queues for AI agents
Push thousands of items at an agent and let the sweep dispatch each one as its own session, at a concurrency you set. A long job stops being one enormous run.
Copy the setup prompt.
Your agent does the rest.
Fetch and execute the appropriate instructions to set me up for superagnt_ from https://mcp.superagnt.com/agent-setup/prompt.md» once it is connected, run this so your agent can switch the source on itself.
agnt_tools_enable({ families: ['queues'] })Everything your agent can do with it
agnt_queues_enqueue takes a single item. agnt_queues_enqueue_batch takes up to 500 payloads in one call, appended in order.
Each item is dispatched as a fresh agent session, so a failure on row 400 does not poison the other 4,000.
A queue dispatches up to its concurrency limit in parallel, from 1 to 20 and defaulting to 1, which is how you keep a big job from saturating a rate limit or a budget.
Paused stops the sweep from dispatching new items while in-flight ones finish. Archived is a terminal retire that keeps the record.
agnt_queues_list shows the queues on a workspace, agnt_queues_list_items shows the items and their state.
A queue targets one deployed agent, so items run against a known prompt and tool set rather than whatever is calling.
tool names07as your agent sees them
agnt_queues_enqueueagnt_queues_enqueue_batchagnt_queues_listagnt_queues_list_itemsagnt_queues_createagnt_queues_updateagnt_queues_deleteThree jobs it can start on today
Enrich a list without one giant run
Batch thousands of rows onto a queue. Each row becomes its own session with its own context and its own cost, and you watch progress instead of waiting on a single call that may not finish.
Fan out research
One agent finds 300 companies worth looking at and enqueues them. A second agent works the queue at a concurrency of 5 and writes results to the workspace database.
Backpressure on a rate-limited step
When a downstream vendor caps you, lower the queue concurrency instead of rewriting the job, and pause it entirely when you need to stop.
What comes back
Your agent reads this response directly, with no parsing layer of your own in between.
https://mcp.superagnt.com/mcptool callagnt_queues_enqueue_batch{
"queue_id": "q_2c81f0",
"payloads": [
{ "domain": "meridian-analytics.example" },
{ "domain": "halcyon-freight.example" }
]
}{
"queue_id": "q_2c81f0",
"items": [
{ "item_id": "qi_9a01", "position": 41 },
{ "item_id": "qi_9a02", "position": 42 }
]
}What it costs to run
Platform tools carry no per-tool price. What you pay is the AI and credit usage the work causes, on the same workspace balance every other tool draws from.
Each run is itemized, so you can see which agent spent what and on which call.
how credits work →Before you wire it up
01How do I turn queue tools on?
Enqueue and inspection are in the default tool surface. Creating, updating and deleting a queue is the queues family, which a person enables on the dashboard MCP page rather than from a tool call, because it can commit the workspace to recurring billed runs.
02Is each item billed separately?
Yes. Every dispatched item is a real agent session and costs what that session costs. Concurrency and pause are your controls on spend.
03What happens to items in flight when I pause?
They finish. Pausing only stops the sweep from dispatching anything new.
04Can I delete a queue that still has items?
Deleting is a hard delete and cascades to every item, pending and completed. When you only want to stop the work, pause it, or archive it to retire it for good.
Point your client at https://mcp.superagnt.com/mcp and your agent has this, plus everything else on one balance.