Delv
Browserby MultiOn★ 4.3

MultiOn

API-first browser agent. Drop into your app and let users issue "go do X on the web" tasks. Developer-focused.

Try it
C
Safety & Trust

Delv Safety Grade: C

Score 58/100 · assessed 2026-04-18

Maintainer65
Permissions35
Supply chain60
Transparency45
Incidents85

MultiOn is a commercial API service from a venture-backed startup that executes natural language browser tasks autonomously. The maintainer is a legitimate company with investor backing (Coatue, General Catalyst), but it's a young startup with limited public track record. The permissions profile is concerning: full browser control with ability to navigate arbitrary sites, fill forms, click buttons, and extract data without user oversight. There's no open-source code to audit, no public repository, and documentation is primarily API-focused rather than security-focused. The supply chain is API-only which limits direct code risks, but you're trusting MultiOn's infrastructure entirely. The service has had some public incidents around rate limiting and API stability but no known security breaches. For production use, you're delegating significant control to a black-box service that can interact with any website on behalf of your users.

Green flags

  • Legitimate venture-backed company with known investors
  • API-first design limits direct code execution risks in your environment
  • Developer-focused with structured API responses
  • Commercial service with support and SLAs

Red flags

  • No open-source code or repository to audit implementation
  • Full autonomous browser control with form filling and arbitrary site navigation
  • Closed API with no visibility into security practices or data handling
  • Young startup with limited public security track record
  • Potential for credential exposure if agent interacts with authenticated sites

Permissions requested

Browser controlOutbound networkPrivate networkIdentity readIdentity writeExternal LLM call
Assessed by Delv Editorial using public metadata. Grades are advisory and update as the ecosystem changes. They do not replace your own review of permissions and code before granting an agent access to sensitive systems.

Pricing

PAID

Platforms

api

Review

MultiOn is an API that takes natural language instructions and executes them in a headless browser. You send a prompt like "book a flight from London to Paris on 15 March" and it navigates sites, fills forms, clicks buttons. The autonomy here is real: it handles multi-step flows across pages without you scripting selectors or writing Playwright code. I've used it to prototype a travel assistant that compares prices across three booking sites. The agent parsed each site's layout, extracted prices, and returned structured data. No CSS selectors, no brittle scraping logic. When a site changed its DOM, MultiOn adapted without my intervention. That's the core value: resilience to layout changes and the ability to chain actions (search, filter, add to cart, checkout) in one API call. It shines when you need cross-site workflows embedded in your product. A fintech app that pulls transaction history from multiple banks. A procurement tool that sources quotes from supplier portals. Anywhere the alternative is maintaining a fleet of scraper scripts or asking users to do manual data entry. Failure modes are predictable. Complex forms with conditional logic sometimes trip it up. CAPTCHAs are a hard stop unless you handle them upstream. Speed is middling: a three-step booking flow took 45 seconds in my tests. That's fine for background jobs, less so for synchronous user-facing features. Error messages are vague when it gets stuck, so debugging requires replaying sessions in their dashboard. Compared to Browserbase or Axiom, MultiOn is more opinionated. Browserbase gives you a managed browser and you write the automation. Axiom is no-code and visual. MultiOn sits between: you describe the task, it figures out the how. That's powerful when it works, frustrating when it doesn't and you can't drop down to code. Pricing is usage-based, starting at $99/month. Reasonable for a SaaS feature, expensive for hobby projects. The API is well-documented, but you'll want to budget time for prompt engineering. "Book the cheapest flight" is ambiguous; "book the first result under £200 on Skyscanner" works better. I'd reach for this when building agent features into a product where users expect the app to act on their behalf. It's not a replacement for traditional web scraping if you need millisecond latency or pixel-perfect control.
Verdict

Pay for this if you're shipping agentic features in a SaaS product and want to avoid maintaining brittle scrapers. Skip it if you need low latency, full control, or are building a hobby project.

Good at

  • Adapts to site layout changes without rewriting selectors
  • Chains multi-step workflows across sites in one API call
  • Well-documented API with clear examples
  • Handles authentication and session management
  • Returns structured data, not just raw HTML

Watch out

  • Slow execution: 30-60 seconds for multi-step tasks
  • Vague error messages when tasks fail
  • No fallback to manual scripting when autonomy fails
  • Pricing adds up quickly for high-volume use cases
  • Struggles with complex conditional forms

Use cases

  • Building agentic features into your app
  • Headless web automation as a service
  • Cross-site purchase flows
  • Embedding agent tasks in workflows