Automation

Parallel browsing, playbook templates, deduplication tracking, workflows, and DAG orchestration

Automation tools handle multi-page and multi-step operations. Use swarm tools for local parallel browsing, playbooks for repeatable automation templates, deduplication tracking to avoid processing the same URL twice, workflows for recording and replaying browser sessions, and DAGs for complex task orchestration with dependencies.

Parallel Browsing

Local parallelism tools that visit multiple URLs concurrently using separate browser instances on your machine.

Tool	Description	Key Parameters
`swarm_fan_out`	Visit multiple URLs in parallel and collect results	`urls` (required), `max_concurrent`
`swarm_collect`	Collect results from a parallel browsing operation	(none)

Playbook Templates

Pre-authored automation templates for common multi-step flows. Playbooks are defined in YAML and support variables, conditionals, loops, and error handling.

Tool	Description	Key Parameters
`swarm_run_playbook`	Execute a YAML playbook or built-in template	`playbook` (required), `variables`, `url`
`swarm_list_playbooks`	List all available playbook templates	(none)
`swarm_playbook_status`	Check progress of a running playbook	(none)

Built-in templates:

greenhouse-apply -- Greenhouse application flow
ashby-apply -- Ashby application flow
lever-apply -- Lever application flow
indeed-apply -- Indeed Easy Apply flow

Playbook step types include: Navigate, Click, Fill, Select, Wait, Extract, Verify, Screenshot, CustomDropdown, UploadFile, SubmitForm, EvalJs, Conditional (if_url_contains, if_variable), and Repeat (for-each loops). Each step can define error handling: Abort, Skip, Retry (with count/delay), or Screenshot.

Deduplication & Verification

SQLite-backed tracking to avoid processing the same URL twice. Records submissions with metadata and verifies success.

Tool	Description	Key Parameters
`swarm_dedup_check`	Check if a URL has already been processed	`url` (required)
`swarm_dedup_record`	Record that a submission was completed	`url` (required), `company`, `title`, `platform`
`swarm_dedup_stats`	Show aggregate deduplication statistics	(none)
`swarm_verify_submission`	Check page for success/error indicators	(none)

Workflow Recording & Replay

Record multi-step browser sessions and replay them with variable substitution.

Tool	Description	Key Parameters
`workflow_start_recording`	Begin recording a replayable workflow	`name` (required)
`workflow_stop_recording`	Stop recording and save the workflow	`description` (required)
`workflow_replay`	Replay a workflow with variable substitution	`name` (required), `variables`
`workflow_list`	List all saved workflows	(none)

DAG Orchestration

Declarative task graphs with dependency management for complex parallel execution.

Tool	Description	Key Parameters
`dag_create`	Create a new task DAG	`name` (required)
`dag_add_task`	Add a task node	`task_id` (required), `description` (required), `action_type` (required), `target`
`dag_add_dependency`	Add a dependency edge between tasks	`task_id` (required), `depends_on` (required)
`dag_ready`	Get tasks ready to execute (all dependencies met)	(none)
`dag_complete`	Mark a task as completed	`task_id` (required), `result` (required)
`dag_progress`	Show DAG completion progress	(none)
`dag_visualize`	Generate a Mermaid diagram of the DAG	(none)

Examples

Fan out across multiple pages

{
  "tool": "swarm_fan_out",
  "arguments": {
    "urls": [
      "https://example.com/page-1",
      "https://example.com/page-2",
      "https://example.com/page-3",
      "https://example.com/page-4"
    ],
    "max_concurrent": 4
  }
}

Visits all URLs in parallel using separate local browser instances, then use swarm_collect to gather the results. The max_concurrent parameter controls how many pages are fetched simultaneously.

Run a playbook template

{
  "tool": "swarm_run_playbook",
  "arguments": {
    "playbook": "greenhouse-apply",
    "variables": {
      "resume_path": "/home/user/resume.pdf",
      "first_name": "Jane",
      "last_name": "Doe",
      "email": "jane@example.com"
    },
    "url": "https://boards.greenhouse.io/company/jobs/12345"
  }
}

Executes the automation template with variable substitution. Check progress with swarm_playbook_status.

Record and replay a workflow

{
  "tool": "workflow_start_recording",
  "arguments": {
    "name": "search-and-extract"
  }
}

All subsequent browse_* tool calls are captured as workflow steps. Stop recording, then replay later with different variables:

{
  "tool": "workflow_replay",
  "arguments": {
    "name": "search-and-extract",
    "variables": {
      "query": "Rust Developer",
      "location": "Remote"
    }
  }
}

Automation

On this page