AstroBrowse - Authenticated Agentic Browser

Core Utility

Coming Soon

Available ActionsEach successful request consumes credits as outlined below.

list_accountsget_policyinitialize_browserrun_stepsextract_pagescreenshotupload_filelist_downloadsdownload_filestart_recordingstop_recordingrequest_user_takeoverwait_for_takeoverstatusclose_browser

Details

Give your AI agent access to the websites and accounts you choose, without ever handing over your passwords. AstroBrowse runs a real, logged-in web browser on your behalf so your agent can work inside the tools you already use: social media, CRMs, ERPs, dashboards, and any SaaS platform or website, including the many that have no API. Save a login once and your agent can resume that session to post content, pull reports, update records, fill forms, download files, and complete multi-step tasks, all in a fresh, single-use browser that is wiped clean when the job is done. If a site ever needs a human touch, like a sign-in confirmation or a quick verification step, AstroBrowse hands you a link to step in for a moment and then lets your agent carry on. You stay in control of which accounts are connected and what the agent is allowed to visit.

Request Beta Access

Beta users get first access to new tools and workflows and help shape the platform. Request access and we will contact you to discuss!

Use Cases

Post and schedule content across social media accounts, Pull reports and update records in CRMs and ERPs, Operate SaaS platforms and portals that have no API, Fill and submit web forms on the user's behalf, Download invoices statements and files from web apps, Log in to accounts without sharing passwords with the agent, Automate multi-step workflows on logged-in websites, Capture screenshots and screen recordings of web sessions, Hand off to a human for sign-in or verification steps, Extract data from pages behind a login

Dynamic MCP Setup

Connect once through AgentPMT Dynamic MCP, then use approved tools from the same agent connection.

30 Second Setup

STDIO connector for Claude Code, Codex, Cursor, Zed, and other LLMs that require STDIO or custom connections.

npm install -g @agentpmt/mcp-routeragentpmt-setup

Hosted Streamable HTTPS

MCP endpoint for browser-based apps like ChatGPT, Claude, Grok, or any time you want a streamable connection with no local install.

https://api.agentpmt.com/mcp

Config Example

Use the hosted endpoint directly in clients that support remote MCP. Store your Bearer token in the client config or secret field.

Full connection guide

{
  "mcpServers": {
    "agentpmt": {
      "type": "streamable-http",
      "url": "https://api.agentpmt.com/mcp",
      "headers": {
        "Authorization": "Bearer <AGENTPMT_BEARER_TOKEN>",
        "x-instance-metadata": "{\"client\":\"generic-mcp\",\"platform\":\"remote\"}"
      }
    }
  }
}

Need client videos, organization controls, audit details, and the full feature overview?

More About Dynamic MCP

Actions(15)

list_accounts0 params

List the user's saved AstroBrowse logins (accounts). Call this first to find a saved site before initialize_browser.

No parameters for this action.

get_policy0 params

Read the user's browsing policy (saved-only vs general browsing). The policy is set only by the human from their dashboard; there is no agent action to change it.

No parameters for this action.

initialize_browser4 params(1 required)

Start a fresh, single-use, isolated browser session. Pass a stable idempotency_key (required) so a retry never starts a second session. Provide account_id to resume a saved login, or omit account_id for a general-browsing session (allowed only when the user has enabled general browsing). Call list_accounts first.

idempotency_keyrequiredstring

Caller-supplied idempotency key for initialize_browser (the agent/request id).

Default: null

account_idstring

Saved account id for an account-backed session. Omit it to start a general-browsing session (no saved login).

Default: null

initial_urlstring

Optional URL to navigate to after initialization. Must match the account's allowed origins.

Default: null

regionstring

Optional region override for a general-browsing session.

Default: null

run_steps2 params(2 required)

Run bounded browser automation steps (goto/click/fill/press/select/wait/extract/screenshot) in an initialized session.

browser_session_idrequiredstring

Opaque browser runtime session id returned by initialize_browser.

Default: null

stepsrequiredarray

Bounded browser automation steps for run_steps.

Default: null

Array of: object

extract_page3 params(1 required)

Extract visible text (or HTML) from the active page or a selector.

browser_session_idrequiredstring

Opaque browser runtime session id returned by initialize_browser.

Default: null

selectorstring

Optional selector for extract_page or upload_file.

Default: null

include_htmlboolean

Whether extract_page should include capped outer HTML.

Default: false

screenshot1 param(1 required)

Capture a PNG screenshot and save it to your File Manager (requires workflow budget context). The image is NOT returned inline; the response has artifact.file_id — fetch it from the File Manager to view or analyze the screenshot.

browser_session_idrequiredstring

Opaque browser runtime session id returned by initialize_browser.

Default: null

upload_file3 params(3 required)

Attach File Manager files to a visible page <input type=file>. Reveal the input with run_steps first, then pass its selector and file_ids.

browser_session_idrequiredstring

Opaque browser runtime session id returned by initialize_browser.

Default: null

selectorrequiredstring

Optional selector for extract_page or upload_file.

Default: null

file_idsrequiredarray

File Manager file_id values for upload_file.

Default: null

Array of: string

list_downloads1 param(1 required)

List files the browser has downloaded in this session (name, size, type) so you can pick one to save with download_file.

browser_session_idrequiredstring

Opaque browser runtime session id returned by initialize_browser.

Default: null

download_file2 params(1 required)

Persist a file the browser downloaded into the File Manager (size-capped, requires workflow budget context). Pass download_name from list_downloads, or omit it to save the most recent download.

browser_session_idrequiredstring

Opaque browser runtime session id returned by initialize_browser.

Default: null

download_namestring

For download_file: the download filename to persist (from list_downloads). Omit to persist the most recent completed download.

Default: null

start_recording2 params(1 required)

Start an MP4 screen recording of the session. show_cursor controls whether the cursor appears in the video.

browser_session_idrequiredstring

Opaque browser runtime session id returned by initialize_browser.

Default: null

show_cursorboolean

For start_recording: capture the real cursor in the video. Only human (takeover) interaction moves the real cursor; agent-driven motion uses the run_steps tutorial fake cursor instead.

Default: true

stop_recording1 param(1 required)

Stop the recording and save the MP4 to your File Manager (requires workflow budget context). The video is NOT returned inline; the response has artifact.file_id — fetch it from the File Manager to view the recording. Call before close_browser.

browser_session_idrequiredstring

Opaque browser runtime session id returned by initialize_browser.

Default: null

request_user_takeover2 params(2 required)

Ask the human to take over the live browser (e.g. CAPTCHA, MFA, unusual login UI). Holds the session open past the idle timeout. Use ONLY when the agent cannot proceed automatically.

browser_session_idrequiredstring

Opaque browser runtime session id returned by initialize_browser.

Default: null

reasonrequiredstring

Reason to show the user when requesting browser takeover.

Default: null

wait_for_takeover1 param(1 required)

Poll the session status after request_user_takeover.

browser_session_idrequiredstring

Opaque browser runtime session id returned by initialize_browser.

Default: null

status1 param(1 required)

Return the session's sanitized live status (no cookies).

browser_session_idrequiredstring

Opaque browser runtime session id returned by initialize_browser.

Default: null

close_browser1 param(1 required)

Release the session: it is wiped and destroyed. Always call this when finished.

browser_session_idrequiredstring

Opaque browser runtime session id returned by initialize_browser.

Default: null

curl -X POST "https://api.agentpmt.com/products/purchase" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ********" \
  -d '{
    "product_id": "6a3ef573573a1ae0a7df8bce",
    "parameters": {
      "action": "list_accounts"
    }
  }'

import requests
import json

url = "https://api.agentpmt.com/products/purchase"

headers = {
    "Content-Type": "application/json",
    "Authorization": "Bearer ********"
}

data = {
    "product_id": "6a3ef573573a1ae0a7df8bce",
    "parameters": {
        "action": "list_accounts"
    }
}

response = requests.post(url, headers=headers, json=data)
print(response.status_code)
print(response.json())

const url = "https://api.agentpmt.com/products/purchase";

const headers = {
  "Content-Type": "application/json",
  "Authorization": "Bearer ********"
};

const data = {
  product_id: "6a3ef573573a1ae0a7df8bce",
  parameters: {
    "action": "list_accounts"
  }
};

fetch(url, {
  method: "POST",
  headers,
  body: JSON.stringify(data)
})
  .then(response => response.json())
  .then(data => console.log(data))
  .catch(error => console.error("Error:", error));

const axios = require('axios');

const url = "https://api.agentpmt.com/products/purchase";

const headers = {
  "Content-Type": "application/json",
  "Authorization": "Bearer ********"
};

const data = {
  product_id: "6a3ef573573a1ae0a7df8bce",
  parameters: {
    "action": "list_accounts"
  }
};

axios.post(url, data, { headers })
  .then(response => {
    console.log(response.status);
    console.log(response.data);
  })
  .catch(error => {
    console.error("Error:", error.message);
  });

Login to view your API and budget keys. The example above uses placeholder values. Sign in to see personalized code with your bearer token.

Usage Instructions

Usage guidance provided directly by the developer for this product.

AstroBrowse - Authenticated Agentic Browser

Operate a real, logged-in web browser on the user's behalf. The user saves their logins once; you resume those sessions to work inside their accounts without ever seeing a password. Every session is single-use and isolated, and is wiped when you close it.

Full setup guide (Tailscale network egress, enabling the tool, saving a login): https://www.agentpmt.com/docs/tool-specific/astrobrowse

Workflow

list_accounts to see the user's saved logins.
initialize_browser with a stable idempotency_key (required, so a retry never starts a second session) plus account_id to resume a saved login, or omit account_id for a general-browsing session when the user's policy allows it. Returns a browser_session_id.
Operate the page with run_steps, extract_page, screenshot, start_recording / stop_recording, upload_file, and list_downloads / download_file.
If a site needs a human (CAPTCHA, MFA, or an unusual login), call request_user_takeover then poll wait_for_takeover. The user steps in on the live session.
Always close_browser when finished.

get_policy tells you whether general browsing is allowed (saved_only | open | open_with_allowlist). Only the user can change the policy, from their dashboard; there is no agent action to change it.

Actions

list_accounts — list saved logins. Call this first.
get_policy — read the browsing policy. It is set only by the user from their dashboard.
initialize_browser — required idempotency_key; optional account_id, initial_url, region. Omit account_id for a general-browsing session (allowed only when the user has enabled it).
run_steps (browser_session_id, steps[]) — bounded steps: goto, click, fill, press, select, wait_for_load_state, wait_for_selector, extract_text, extract_html, evaluate_readonly, screenshot. Each step takes optional selector, text, url, key, value, script, timeout_ms (500 to 60000).
extract_page (browser_session_id, selector?, include_html?) — visible text or capped HTML.
screenshot (browser_session_id) — PNG saved to the File Manager.
upload_file (browser_session_id, selector, file_ids[]) — attach File Manager files to a visible <input type=file>. Reveal the input with run_steps first.
list_downloads / download_file (browser_session_id, download_name?) — list files the page downloaded and persist one to the File Manager.
start_recording (show_cursor?) / stop_recording — MP4 of the session saved to the File Manager. Stop before close.
request_user_takeover (reason) / wait_for_takeover — hand the live session to the human, then poll for status.
status — sanitized live status (no cookies).
close_browser — release and wipe the session. Always call when done.

Notes

Artifact actions (screenshot, download_file, stop_recording, upload_file) run inside a workflow with budget context.
Credentials are never exposed to the agent.
Sessions are single-use and isolated; closing wipes them.
Full setup and configuration guide (Tailscale egress, enabling the tool, saving a login): https://www.agentpmt.com/docs/tool-specific/astrobrowse

About this Product

Give your agent access to your accounts, never your passwords

AstroBrowse lets your AI agent work inside the websites and accounts you already use, without ever handing over your credentials. Save a login once and your agent can resume that session in a real, logged-in browser to run the workflows you assign, from posting on social media to updating your CRM, pulling reports from an ERP, or operating any SaaS platform or website, including the many that have no API at all.

How it works

You choose the accounts. Connect the logins you want your agent to use. The agent operates those sessions but never sees your passwords.
Your agent does the work. It navigates, clicks, fills forms, uploads and downloads files, extracts data, and captures screenshots or screen recordings as proof.
You step in only when needed. If a site asks for a sign-in confirmation or a quick verification, AstroBrowse sends you a link to take over the live browser for a moment, then your agent picks right back up.
Every session is clean. Each task runs in a fresh, single-use browser that is wiped when the job is done.

What you can automate

Social media: draft, publish, and schedule posts across your brand accounts.
CRMs and sales tools: update records, log activity, and export pipeline reports.
ERPs and back-office systems: enter data, download statements, and run routine operations.
Any SaaS or website: operate portals and dashboards that have no API, on your behalf.
Document workflows: upload files into web forms and save downloads straight to your files.

You stay in control

Decide exactly which accounts your agent can use and where it is allowed to browse, with a policy that ranges from saved logins only to open browsing within an allowlist. Screenshots and recordings give you a clear record of everything the agent did. It is the simplest way to put your agent to work across the tools you rely on, safely and on your terms.

Full setup instructions

New to AstroBrowse? The complete setup guide walks you through connecting your Tailscale network egress, enabling AstroBrowse for an agent, and saving your first login: AstroBrowse setup guide.

Frequently Asked Questions

How do I connect this tool to an external agent?

You can install the local MCP server by opening a terminal and running:

Install commands

npm install -g @agentpmt/mcp-router
agentpmt-setup

This will connect you to local agents like Claude Code, Windsurf, Grok Build, Cursor, etc.

Alternatively you can connect to the hosted version with this config block, no installation required:

Hosted MCP config

{
  "mcpServers": {
    "agentpmt": {
      "type": "streamable-http",
      "url": "https://api.agentpmt.com/mcp",
      "headers": {
        "Authorization": "Bearer <AGENTPMT_BEARER_TOKEN>",
        "x-instance-metadata": "{\"client\":\"generic-mcp\",\"platform\":\"remote\"}"
      }
    }
  }
}

View MCP Connection Instructions for more details.

How does an external agent use this tool?

After the external agent is connected to an Agent Group that can use this tool, paste this prompt into the agent:

Agent prompt

Use the AgentPMT-Tool-Search-and-Execution tool. First call action 'get_instructions' so you know how to use the tool search interface. Then call action 'get_schema' with tool_id 6a3ef573573a1ae0a7df8bce ("AstroBrowse - Authenticated Agentic Browser"). After reading the schema and any returned instructions, tell me what this tool can do, we are going to be using it

The agent should fetch the tool schema first, collect the required parameters for your request, and then call the tool through AgentPMT.

Can it save files and screenshots for me?

Yes. Screenshots, screen recordings, and any files the page downloads are saved to your File Manager so you can use them in the rest of your workflow.

Do I control where the agent can go?

Yes. You set a browsing policy that can restrict the agent to your saved logins only, or allow broader browsing within an allowlist you define.

Does my agent see my passwords?

No. You save your logins once, and AstroBrowse resumes those sessions for your agent. The agent operates the logged-in browser but never has access to your credentials.

Is each session private and temporary?

Yes. Every task runs in a fresh, single-use browser that is wiped when the session closes, so nothing carries over between jobs.

What can the agent actually do in the browser?

It can navigate, click, fill and submit forms, select options, extract text and data, upload files, download files, and capture screenshots or full screen recordings of the session.

What happens if a site needs a human, like a CAPTCHA or a verification code?

AstroBrowse sends you a link to take over the live browser for a moment. Once you have handled the step, your agent continues automatically.

What sites and accounts can it use?

Any website or account you connect, including social media, CRMs, ERPs, dashboards, and SaaS platforms. It is especially useful for tools that have no API, because it works through the real web interface.

AstroBrowse - Authenticated Agentic Browser

Available ActionsEach successful request consumes credits as outlined below.

Details

Request Beta Access

Use Cases

Dynamic MCP Setup

30 Second Setup

Hosted Streamable HTTPS

Config Example

Actions(15)

Credit-Based Access Using AgentAddress

Direct x402 Payment

Usage Instructions

AstroBrowse - Authenticated Agentic Browser

Workflow

Actions

Notes

About this Product

Give your agent access to your accounts, never your passwords

How it works

What you can automate

You stay in control

Full setup instructions

Frequently Asked Questions

How do I connect this tool to an external agent?

How does an external agent use this tool?

Can it save files and screenshots for me?

Do I control where the agent can go?

Does my agent see my passwords?

Is each session private and temporary?

What can the agent actually do in the browser?

What happens if a site needs a human, like a CAPTCHA or a verification code?

What sites and accounts can it use?

Looking for help integrating AI into your business? Set up a free consultation.