Run once without an automation

One-off operations create a run or scrape job without creating a cron schedule or dashboard automation record.

One-off operation matrix

Capability	Start operation	Status operation	Creates an automation
Scrape a URL	POST /scrape	GET /scrape/status/{jobId}	No
Agentic Scraper browser agent	POST /agenticscraper	GET /agenticscraper/status/{jobId}	No
Custom agent	POST /agents/run	GET /agents/runs/{jobId}	No
Agent Market agent	POST /agent-market/agents/{agentId}/run	GET /agent-market/runs/{jobId}	No
Reddit agent	No standalone public one-off operation	—	—
X agent	No standalone public one-off operation	—	—

Do not use POST /automations as a one-off substitute. It immediately queues a first execution, but it also stores a recurring automation. The public REST API does not currently expose a standalone Reddit/X one-off operation.

Scrape a URL asynchronously

curl --request POST \
  --url https://api.agenticscraper.com/api/v1/scrape \
  --header "Content-Type: application/json" \
  --header "X-API-Key: $AGENTIC_SCRAPER_API_KEY" \
  --data '{
    "url": "https://example.com/products",
    "async": true,
    "structured_output": [
      { "id": "product_names", "type": "string[]" },
      { "id": "prices", "type": "string[]" }
    ]
  }'

Poll the returned job:

curl \
  --url "https://api.agenticscraper.com/api/v1/scrape/status/$JOB_ID" \
  --header "X-API-Key: $AGENTIC_SCRAPER_API_KEY"

Set "async": false to keep the request open and return the scrape result directly.

Run Agentic Scraper once

Use this when the task requires an autonomous browser session:

curl --request POST \
  --url https://api.agenticscraper.com/api/v1/agenticscraper \
  --header "Content-Type: application/json" \
  --header "X-API-Key: $AGENTIC_SCRAPER_API_KEY" \
  --data '{
    "task": "Visit the target site, compare the three pricing plans, and return their limits.",
    "llm_provider": "agentic_scraper",
    "llm_model": "google/gemini-2.5-flash",
    "vision_mode": "auto",
    "optimization_mode": "balanced",
    "max_actions_per_step": 8,
    "structured_output": [
      { "id": "plans", "type": "string[]" },
      { "id": "summary", "type": "string" }
    ],
    "headless": true
  }'

Poll:

curl \
  --url "https://api.agenticscraper.com/api/v1/agenticscraper/status/$JOB_ID" \
  --header "X-API-Key: $AGENTIC_SCRAPER_API_KEY"

Run a custom agent once

This endpoint accepts the custom-agent payload without storing an automation:

curl --request POST \
  --url https://api.agenticscraper.com/api/v1/agents/run \
  --header "Content-Type: application/json" \
  --header "X-API-Key: $AGENTIC_SCRAPER_API_KEY" \
  --data '{
    "name": "One-off documentation researcher",
    "task": "Use the selected documentation tool to explain the provider event charging flow.",
    "system_prompt": "Use tools before answering and cite the retrieved documentation.",
    "llm_provider": "agentic_scraper",
    "llm_model": "deepseek/deepseek-chat-v3-0324",
    "max_tool_calls": 10,
    "market_mcps": [
      {
        "mcp_id": "<approved-mcp-uuid>",
        "tool_names": ["search_documentation", "get_document"],
        "params": {}
      }
    ],
    "structured_output": [
      { "id": "answer", "type": "string" },
      { "id": "sources", "type": "string[]" }
    ]
  }'

Poll:

curl \
  --url "https://api.agenticscraper.com/api/v1/agents/runs/$JOB_ID" \
  --header "X-API-Key: $AGENTIC_SCRAPER_API_KEY"

max_tool_calls bounds the whole execution, including paid MCP calls.

Run an Agent Market agent once

First list the approved catalog and obtain an agent UUID:

curl \
  --url https://api.agenticscraper.com/api/v1/agent-market/agents \
  --header "X-API-Key: $AGENTIC_SCRAPER_API_KEY"

Run a task-mode agent:

curl --request POST \
  --url "https://api.agenticscraper.com/api/v1/agent-market/agents/$AGENT_ID/run" \
  --header "Content-Type: application/json" \
  --header "X-API-Key: $AGENTIC_SCRAPER_API_KEY" \
  --data '{
    "task": "Research the requested topic and return a source-backed summary.",
    "llm_provider": "agentic_scraper",
    "llm_model": "google/gemini-2.5-flash",
    "webhook_url": "https://example.com/webhooks/agentic-scraper",
    "headless": true
  }'

For a template-mode agent, send keys declared by its inputSchema:

{
  "inputs": {
    "topic": "browser agents",
    "maximumResults": 10
  }
}

Poll:

curl \
  --url "https://api.agenticscraper.com/api/v1/agent-market/runs/$JOB_ID" \
  --header "X-API-Key: $AGENTIC_SCRAPER_API_KEY"

Purchased/accessible Agent Market rules, wallet balance, provider tool charges, LLM usage, and browser runtime usage still apply to one-off executions.

Polling safely

All asynchronous job records use terminal states completed and failed. Poll with exponential backoff, stop on a terminal state, and never assume a failed run is free: upstream LLM, tool, or browser resources consumed before the failure may still be charged.