Claim prioritized work and preview the dispatch queue using the Loop TypeScript SDK.

Dispatch

The dispatch resource is Loop's core agent feedback loop. It lets agents claim the highest-priority unblocked issue with hydrated prompt instructions, or preview the queue without claiming anything.

Access it via loop.dispatch on your client instance.

import { LoopClient } from '@dork-labs/loop-sdk'

const loop = new LoopClient({ apiKey: 'loop_abc123' })

next()

Atomically claim the highest-priority unblocked issue. Returns the issue with hydrated prompt instructions, or null if the queue is empty.

const task = await loop.dispatch.next()

if (task) {
  console.log(task.issue.title)
  console.log(task.prompt)
}

Parameters

Parameter	Type	Required	Description
`projectId`	`string`	No	Filter by project ID

Return type

Promise<DispatchNextResponse | null>

Returns null when no unblocked issues are available (the API returns 204 No Content).

DispatchNextResponse

Field	Type	Description
`issue`	`object`	Claimed issue with `id`, `number`, `title`, `type`, `priority`, `status`
`prompt`	`string \| null`	Fully hydrated prompt instructions for the agent
`meta`	`object \| null`	Template metadata (see below)

The meta object contains:

Field	Type	Description
`templateSlug`	`string`	Slug of the matched prompt template
`templateId`	`string`	Template ID
`versionId`	`string`	Active version ID
`versionNumber`	`number`	Active version number
`reviewUrl`	`string`	URL for submitting a prompt review

Filter by project

const task = await loop.dispatch.next({ projectId: 'clx1abc2def3ghi4jkl5mnop' })

queue()

Preview the priority queue without claiming any issues. Returns a paginated list of unblocked todo issues with their score breakdowns.

const queue = await loop.dispatch.queue()

for (const item of queue.data) {
  console.log(`#${item.issue.number} ${item.issue.title} — score: ${item.score}`)
}

Parameters

Parameter	Type	Required	Default	Description
`projectId`	`string`	No	none	Filter by project ID
`limit`	`number`	No	none	Maximum items per page
`offset`	`number`	No	`0`	Number of items to skip

Return type

Promise<PaginatedList<DispatchQueueItem>>

The PaginatedList wrapper provides data, total, and a hasMore getter for pagination.

DispatchQueueItem

Field	Type	Description
`issue`	`Issue`	Full issue object
`score`	`number`	Composite priority score
`breakdown`	`object`	Score breakdown (see below)

The breakdown object contains:

Field	Type	Description
`priorityWeight`	`number`	Weight based on priority level
`goalBonus`	`number`	Bonus for issues linked to a project with a goal
`ageBonus`	`number`	Bonus that increases as the issue ages
`typeBonus`	`number`	Bonus based on issue type

Paginated example

const queue = await loop.dispatch.queue({ limit: 5 })

console.log(`Showing ${queue.data.length} of ${queue.total} items`)
console.log(`Has more: ${queue.hasMore}`)

Agent polling loop

A typical agent integration polls next() on a schedule, executes the prompt, reports results, and repeats.

import { LoopClient } from '@dork-labs/loop-sdk'

const loop = new LoopClient({ apiKey: process.env.LOOP_API_KEY! })

async function agentLoop() {
  const task = await loop.dispatch.next()

  if (!task) {
    console.log('No work available')
    return
  }

  console.log(`Claimed #${task.issue.number}: ${task.issue.title}`)

  // Execute the prompt instructions with your AI model
  const result = await executeWithAgent(task.prompt)

  // Report results back to Loop
  await loop.issues.update(task.issue.id, {
    status: 'done',
  })
}

// Poll every 60 seconds
setInterval(agentLoop, 60_000)

Dispatch concepts for priority scoring, blocking filters, and SKIP LOCKED concurrency
Templates for managing the prompt templates that dispatch hydrates
Issues for updating issues after completing dispatched work

Dispatch

On this page