Parallel

Web research for AI agents: live web search, page extraction and deep research task runs from Parallel.

5 actions API key 4 read1 write Calls api.parallel.ai

Connect

TypeScript

const conn = await setu.connections.initiate({ userId: "user_1", toolkit: "parallel" });
console.log(conn.connectUrl); // show this to your user
await conn.waitUntilActive();

Permissions requested

This toolkit uses a single credential and requests no scopes.

Actions

Start research taskwriteparallel.create_task_runStart an asynchronous Parallel research task from a question or JSON input. Returns a run_id immediately; fetch the answer later with get_task_result. Uses credits.

Input	Type		Description
`input`	string	Required	The question or task, as text
`processor`	string	Required	Processor tier, for example base or core (higher tiers research deeper and cost more)
`task_spec`	object	Optional	Optional output spec, for example {"output_schema":"a one-line answer"}

TypeScript

await setu.execute({
  userId: "user_1",
  action: "parallel.create_task_run",
  input: {
    "input": "<input>",
    "processor": "<processor>"
  },
});

Extract page contentreadparallel.extractFetch up to 20 web pages and return their content as markdown excerpts, optionally focused on an objective.

Input	Type		Description
`urls`	array of string	Required	Page URLs to read (up to 20)
`objective`	string	Optional	What to focus the excerpts on
`max_chars_total`	integer	Optional	Upper bound on characters across all excerpts

TypeScript

await setu.execute({
  userId: "user_1",
  action: "parallel.extract",
  input: {
    "urls": []
  },
});

Get task resultreadparallel.get_task_resultGet the finished output of a Parallel task run, waiting up to the timeout for it to complete. A 408 reply means it is still running: call again.

Input	Type		Description
`run_id`	string	Required	Run ID returned by create_task_run
`timeout`	integer	Optional	Seconds to wait for completion (default 600)

TypeScript

await setu.execute({
  userId: "user_1",
  action: "parallel.get_task_result",
  input: {
    "run_id": "<run_id>"
  },
});

Get task run statusreadparallel.get_task_runCheck the status of a Parallel task run (queued, running, completed, failed) without waiting for its result.

Input	Type		Description
`run_id`	string	Required	Run ID returned by create_task_run

TypeScript

await setu.execute({
  userId: "user_1",
  action: "parallel.get_task_run",
  input: {
    "run_id": "<run_id>"
  },
});

Search the webreadparallel.searchSearch the live web and get ranked results with URLs, titles and relevant excerpts. Pass an objective and 2-3 short keyword queries.

Input	Type		Description
`objective`	string	Optional	What you are trying to find, in plain language, self-contained
`search_queries`	array of string	Required	Keyword queries, 3-6 words each; 2-3 work best
`mode`	string	Optional	Speed and depth trade-off (default advanced)
`max_chars_total`	integer	Optional	Upper bound on characters across all excerpts

TypeScript

await setu.execute({
  userId: "user_1",
  action: "parallel.search",
  input: {
    "search_queries": []
  },
});