NEWAI-assisted RPA with review and approval →

Automate browser work. Keep every run under control.

Argus is a local-first desktop workspace for teams to build, approve, run and monitor repeatable browser workflows with RPA, MCP and an open Local API.

Windows · macOS · 3 workspaces free · No card required

Local API · localhost:50325
// 1. Launch a profile, get its CDP endpoint
const res = await fetch(
  'http://localhost:50325/api/v2/browser-profile/start',
  { method: 'POST', body: JSON.stringify({ profile_id }) }
).then(r => r.json());

// 2. Drive it with Puppeteer — or hand it to your AI agent
const browser = await puppeteer.connect({
  browserWSEndpoint: res.data.ws.puppeteer,
});

One controlled path from an idea to a verifiable browser run

RPA SDK
Human approval
Local Runner
Schedules
Run timeline
Evidence
Data export
Team access
For operations teams

Turn repeatable browser work into controlled runs

Define the workflow, review exactly what it will do, and send it to the right local Runner. Every run returns a status, timeline and evidence.

Versioned RPA scripts

Build deterministic browser workflows with the current RPA SDK, validate them, and activate one immutable version at a time.

  • Immutable script versions
  • SDK and risk validation
  • Constrained desktop execution

AI-assisted, human-approved

Use MCP to turn a structured goal into a reviewable plan and script. Saving, publishing and running remain separate confirmed actions.

  • Readable execution projection
  • Explicit publish and run approval
  • Exact script version pinned
🖥

Schedule and operate runs

Dispatch work to a selected Runner, schedule approved versions, inspect progress, cancel one task, and diagnose failures from evidence.

  • One-time and recurring schedules
  • Run timelines and evidence
  • Evidence-gated retry diagnostics
👥

Keep output under your control

Send structured results to local files, your own webhook or Google Sheets. Destination details and secrets stay encrypted on the Runner.

  • JSONL, CSV and SQLite
  • Encrypted delivery queue
  • User-controlled destinations
Acceptable use

Argus is for browser workflows you own or are authorised to perform. Spam, fake engagement, credential abuse and attempts to bypass third-party access controls are prohibited.

For AI builders

AI can build the automation. You stay in control.

Argus gives MCP clients a governed path from plan to script to approved run, with explicit confirmation at every side-effect boundary.

🔌

A broad MCP toolset

Connect a compatible MCP client to plan, build, validate and operate automation through documented tools.

🎛

Plan before code

A build session captures the site, steps, inputs, outputs, limits and failure policy before code is saved.

🍪

Immutable versions

Draft, verified and active states are separate. A published version cannot silently change beneath a scheduled run.

🌐

Approve the exact run

Resolve the script version, Runner, workspace, parameters, output destination and limits before execution.

📊

Evidence-backed outcomes

Read structured status, run timelines and failure diagnoses instead of asking an agent to guess what happened.

Optional
📝

Bring your own model

External AI calls are optional and limited to workflows that explicitly use them; provider credentials remain unreadable to RPA scripts.

Developer tools

Open where it matters

Use the RPA SDK, MCP or localhost API. Desktop Runners execute the work; cloud services coordinate versions, schedules and run status.

Local API

Start a Workspace

Launch an authorised browser workspace and get its CDP endpoint from localhost:50325.

curl -X POST http://localhost:50325/api/v2/browser-profile/start \
  -H "Content-Type: application/json" \
  -d '{"profile_id": "..."}'
Puppeteer

Use Your Framework

Connect Puppeteer or Playwright to the returned CDP endpoint for custom local automation.

const browser = await puppeteer.connect({
  browserWSEndpoint: wsEndpoint
});
MCP

Build with AI, Safely

Begin with a structured plan. Argus returns a readable projection for human confirmation before a draft can be saved.

argus_rpa_begin_build
argus_rpa_validate_script
RPA SDK 1.3

Versioned Scripts

Build deterministic workflows in a constrained SDK, validate them, and publish an exact immutable version.

export async function run(ctx, params) {
  // reviewed browser work
}
Output API

Keep the Data

Write structured results to local JSONL, CSV or SQLite, or to a destination configured on your Runner.

await ctx.output.append({
  dataset: 'results', records
});
Run Operations

Inspect Every Run

Read the ordered lifecycle, action, evidence and error timeline, then diagnose or cancel one exact task.

argus_rpa_get_run_timeline
argus_rpa_diagnose_run

Technical documentation for the Local API, MCP and current RPA execution model.

View Technical Documentation
Security & governance

Automation that runs inside clear boundaries

Argus separates planning, approval, execution and evidence. Sensitive browser state stays local or encrypted, while teams control who can build, publish, run and inspect workflows.

  • Explicit control points
    Building, publishing, running and scheduling are separate actions with human confirmation at side-effect boundaries.
  • Local-first execution
    Desktop Runners execute browser work on machines you manage, close to your existing data and authenticated sessions.
  • Encrypted synchronization
    Supported sensitive state is encrypted before sync; local output destinations and secrets remain on the Runner.
  • Permissions and audit trail
    Team roles govern who can manage, publish and run automation, with actions recorded for review.
LOCAL
Managed desktop execution
E2EE
Encrypted sync
RBAC
Role-based access
AUDIT
Traceable operations
Pricing

Transparent pricing — choose what you need

Choose a monthly workspace plan based on browser workspaces, concurrent runs and team access.

Free

Explore local browser automation

USD $0Forever Free
  • 3 browser workspaces
  • 1 concurrent workspace
  • 2 team members
  • Local API and CDP access
  • Free RPA templates
  • Community support
Free Download

Starter

For individual automation builders

USD $19/mo
  • 20 browser workspaces
  • 3 concurrent workspaces
  • 2 team members
  • 80 workspace opens per day
  • Local API and CDP access
  • Shared workspace quota
Get Started
Most Popular

Pro

For teams building governed RPA

USD $49/mo
  • 100 browser workspaces
  • 10 concurrent workspaces
  • 5 team members
  • 300 workspace opens per day
  • Custom versioned RPA scripts
  • MCP-assisted workflow building
Get Started

Business

For growing automation operations

USD $99/mo
  • 300 browser workspaces
  • 30 concurrent workspaces
  • 15 team members
  • 1,000 workspace opens per day
  • Run operations and diagnostics
  • Priority support
Get Started

Enterprise

For large teams

Contact Sales
  • Custom workspace quota
  • Custom concurrency quota
  • Custom team quota
  • Custom member seats
  • Self-hosted deployment
  • Dedicated account manager
Contact Us

Prices shown in USD. Paid subscriptions renew automatically until cancelled. Initial self-service purchases are covered by the 7-day refund policy, subject to its terms and exclusions.

FAQ

Still have questions?

We've collected the most common questions. Didn't find your answer? Get in touch.

Repeatable browser work on systems and accounts you own or are authorised to use: data entry, form processing, QA, internal administration and permitted structured data collection. Build a versioned RPA script, connect through MCP, or use the Local API with Puppeteer or Playwright. Third-party terms and applicable law still apply.

Move repeatable browser work into a controlled workflow

3 workspaces free · No card required · Review before you run