Continue a task

Every run returns a session_id. Pass it into a new run to continue the same thread — the agent keeps the context of what came before. You do not create a session. The first run mints it; you just reference it.

Continue

from browser_use import BrowserUse

client = BrowserUse()

first = client.agents.browser_use.run(task="Search Amazon for a mechanical keyboard")
follow = client.agents.browser_use.run(
    task="Add the top-rated one to the cart",
    session_id=first.session_id,
)
print(follow.result)

import { BrowserUse } from "browser-use";

const client = new BrowserUse();
const first = await client.agents.browserUse.run({ task: "Search Amazon for a mechanical keyboard" });
const follow = await client.agents.browserUse.run({
  task: "Add the top-rated one to the cart",
  session_id: first.session_id,
});
console.log(follow.result);

# first run returns { "sessionId": "ses_xyz789", ... }
curl -X POST https://api.browser-use.com/api/v1/agents/browser-use/runs \
  -H "X-Browser-Use-API-Key: $BROWSER_USE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"task": "Add the top-rated one to the cart", "session_id": "ses_xyz789"}'

On a continuation you send only the new task (plus optional attached_file_ids to add more files). Everything else is inherited from the conversation and you do not re-pass it:

Stay logged in — the browser_profile_id carries over; a follow-up is already authenticated. Don’t re-send it.
Proxy, screen size, model — inherited too. You can’t change them mid-thread; start a new run for a different config.
Files — every file the thread added or produced is still on the conversation’s disk. New files: add + attach them on the follow-up.

Request bodies accept both session_id and sessionId; responses always return sessionId. The list filter query param is ?sessionId= — sending ?session_id= is silently ignored and returns all your runs.

Agent differences

browser-use — the session lives as long as its browser. Once the browser is gone (stopped, or idle-reaped), the session is done: continuing a closed session returns a clean 4xx (“session ended, start a new run”). browser-use is task-scoped by nature.
browsercode — the session outlives its browsers and can run for hours. If a session is busy, a continuation is queued and runs automatically when the current turn finishes (same call, status: "queued").

List the runs in a session

runs = client.agents.browser_use.runs.list(session_id="ses_xyz789")

curl "https://api.browser-use.com/api/v1/agents/browser-use/runs?sessionId=ses_xyz789" \
  -H "X-Browser-Use-API-Key: $BROWSER_USE_API_KEY"

Get started

Agents

Browsers

Account

More

Continue

Agent differences

List the runs in a session

​Continue

​Agent differences

​List the runs in a session

Continue

Agent differences

List the runs in a session