By stemado
Stealth browser automation for websites with anti-detection. Scout page structure, execute actions, monitor network, then compose botasaurus-driver scripts.
Attach to a paused otto-engine execution's browser for Hot Takeover
Run Otto performance benchmarks and produce a versioned results file
Export the current browser session as a replayable workflow (Python script + portable JSON)
Report a bug, request a feature, or flag UX friction as a GitHub issue
Resume a paused otto-engine execution after Hot Takeover
Admin access level
Server config contains admin-level keywords
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Turn any web portal into a programmable API.
No web development knowledge required.
Every enterprise has systems that were never designed to be automated -- vendor dashboards, HR portals, benefits platforms, internal tools locked behind logins with no API in sight. Otto fills that gap.
Otto is a Claude Code plugin. You describe a task conversationally -- "log into the portal, navigate to Reports, export the CSV" -- and Claude scouts the page structure, drives the browser through the workflow, and exports a standalone Python script that runs unattended. It conforms to whatever interface it encounters and produces a replayable automation that behaves like the API that should have existed all along.
| Otto | Dev-Browser | Chrome Extension MCP | Playwright MCP | |
|---|---|---|---|---|
| Works on portals you don't control | Yes -- enterprise portals, vendor dashboards, HR systems | Limited -- designed for your own app | Limited -- your active browser session | Limited -- designed for your own app |
| Page discovery | Compact scout reports (~200 tokens) | You provide selectors | You provide selectors | You provide selectors |
| Credential safety | Never appears in conversation | Plaintext in context | Plaintext in context | Plaintext in context |
| XHR interception | Built-in network monitor with filtering | No | No | Limited |
| Export to script | One command, runs unattended | No | No | No |
| Encrypted credentials | Vault with OS keyring | No | No | No |
| Credential scrubbing | Automatic in all output | No | No | No |
| Video recording | CDP screencast to MP4 | No | No | Built-in |
| Dependency pre-checks | SessionStart hook | No | No | No |
| Hot Takeover | Attach to a paused workflow's live browser, debug interactively, resume | No | No | No |
| Best for | Enterprise systems without APIs | Testing your own app | Interactive debugging | Cross-browser test suites |
Otto's compact scout reports and targeted extraction use dramatically fewer tokens than full-page snapshot approaches.
| Task | Otto | Playwright MCP | Reduction |
|---|---|---|---|
| Fact lookup (Wikipedia) | 1,264 tokens / 4 tool calls | ~124,000 tokens / 1 tool call | 98× fewer |
| Form fill (httpbin) | 3,799 tokens / 11 tool calls | ~124,000+ tokens | 33× fewer |
Wall-clock: 2.2 s (fact lookup) · 8.2 s (form fill) on Sonnet 4.6 — 100% success rate, zero retries across 6 runs.
Full methodology, run-by-run data, and model comparison → Benchmark results
Enterprise portals are tough terrain for automation. The login form lives inside a triple-nested iframe. The export button triggers an XHR that returns a signed URL instead of a file. The page uses dynamic selectors that change between deployments.
Other browser automation tools require you to already know the page structure and provide selectors upfront. Otto takes the opposite approach -- Claude scouts the page, discovers selectors from reconnaissance rather than guesswork, drives the workflow step by step, and exports a production-grade script. No web development knowledge required.
New here? The Getting Started guide walks through a complete example from scout to scheduled script.
The current showcase tells a compact story — scout, automate, export. The next version expands this into a seven-phase cinematic narrative that walks stakeholders from the initial pain ("your data is trapped behind a login screen") through teaching Otto, workflow transformation, delivery, internal discovery, the API switch, and finally integration at scale.
Two design documents define this upcoming work:
npx claudepluginhub stemado/otto-marketplace --plugin ottoInteractive HTML visualization companion — push diagrams, mockups, and explorers to a browser window
Interactive HTML visualization companion — push diagrams, mockups, and explorers to a browser window
Stealth browser automation for websites with anti-detection. Scout page structure, execute actions, monitor network, then compose botasaurus-driver scripts.
AI-powered browser automation -- lets Claude control real web browsers to navigate, click, type, extract content, and automate workflows
Anti-detect browser automation skill using Patchright (undetected Playwright fork)
Actionbook MCP server for up-to-date browser automation. Get verified CSS/XPath selectors for any website directly into your agent context.
Browser automation with persistent page state. Use when users ask to navigate websites, fill forms, take screenshots, extract web data, test web apps, or automate browser workflows.
Universal Self-Learning Browser Agent — record browser interactions, generate API skills, and replay them via MCP tools with progressive tier optimization.
Ultra-fast browser automation with 98% token reduction via batch execution and compact snapshots