Back to vendors

CopyCat AI

Visit site

Browser / computer-use agentindependentVerified 2026-07-01

AI powered robotic process automation that builds no code browser agents combining intelligent agentic steps with precise deterministic actions to automate legacy web portals that lack APIs.

CopyCat is an AI powered robotic process automation platform, built by a Y Combinator team and launched in 2025, that lets people build browser agents to automate web based workflows that traditional automation cannot reach. The founders started it after years of writing brittle browser scripts at their previous company and set out to replace clunky legacy robotic process automation with something that does not rely on fragile scripts, virtual machines, or offshore labor. The platform has powered more than one hundred thousand successful executions and saved customers thousands of hours, and it has increasingly focused on the insurance industry.

Its central idea is a hybrid of two kinds of steps. Deterministic steps like click and type, including precise XPath selectors, handle specific actions reliably, while agentic steps let an AI figure out how to perform dynamic actions such as logging in with a set of credentials or navigating a page whose elements keep changing. Users combine these in a no code visual builder, adding structure steps like if and else and for each loops, so a workflow can be both intelligent where it needs to adapt and precise where it needs to be exact. This directly targets the legacy portals that have no application programming interface.

CopyCat runs automations in the cloud with no virtual machines or desktop installs, and workflows can be triggered on a schedule, through an application programming interface, or by webhooks, with browser proxies available for scale. Customers use it to upload invoices into legacy transportation management systems, download and reconcile bank statements, fill complex healthcare and insurance forms, and extract data from government websites that hold information behind clunky interfaces. Where an interface exists but an integration does not, CopyCat can drive it.

As a young company the platform is still filling in enterprise capabilities. Role based access controls and versioning were described by the founders as near term additions rather than shipped features, and named security certifications, persistent memory, self hosted deployment, and prebuilt agent libraries were not documented. Its clear strength is the intelligent browser navigation at its core, paired with deterministic precision, which suits it to automating the messy legacy web that defeats both classic scripts and pure large language model agents.

Vendor details

Canonical URL

https://runcopycat.com

Company status

independent

Use cases & customers

In practice

A logistics team must upload thousands of invoices into a legacy transportation management system with no API. A CopyCat agent logs in with an agentic step, then uses deterministic actions to upload and file each invoice, accelerating payments.

A brokerage fills the same complex insurance forms by hand every day. They build a CopyCat workflow that navigates the carrier's changing web forms with agentic steps and enters data precisely, processing far more submissions without adding staff.

A real estate firm needs owner details buried in government portals. CopyCat extracts the data at scale, navigating clunky public websites that have no integrations, saving hundreds of hours of manual lookups each month.

Sources & related URLs

Research notes

'AI RPA powered by browser agents.' No-code platform: HYBRID of AGENTIC steps (AI browser agents navigate intelligently, figure out dynamic actions like 'login with these credentials', handle changing elements) + DETERMINISTIC steps (click/type/XPath selectors for precision). Mix both in one workflow + structure steps (If/Else, For Each, Get-List-of-Child-Elements). YC W25. San Francisco, founded 2024. Founders Graham Sabin, Abhi Balijepalli, Zyad Elgohary (ex-Platter, ex-Zoom). 'Replace clunky legacy RPA' (brittle scripts/VMs/offshore-labor). 100,000+ executions, 5000+ hrs saved. Now INSURANCE-focused (brokers/carriers/MGAs). No-code visual builder. Runs CLOUD ('no VMs, no desktop installs'). Triggers: schedule/API/webhooks. Browser proxies. Use cases: invoice-uploads to legacy TMS (logistics), invoice-validation (AppFolio), bank-statement download/reconcile (fintech), insurance/healthcare form-filling, gov-site (real estate). NLP models interpret instructions→browser commands. 1 FULL: COMP=F (browser agents navigate websites INTELLIGENTLY, handle dynamic pages, + deterministic precision — core; genuine AI browser-use, more agentic than Axiom). 7 P's: Int=P (any site + webhooks + API-trigger), Orch=P (multi-step + branching/loops + agentic-steps; not fully-autonomous-goal-driven), Know=P (NLP interprets page context/intent + data-extraction), Obs=P (build/test + watch-runs), Trig=P (schedule/API/webhook), Ext=P (API-trigger + webhooks), Eval=P (design/test/refine + deterministic-reliability). 6 N's: HITL=N (deterministic-constraint is design-time, no runtime human-approval), Sec=N (cloud + proxies; NO named certs, RBAC EXPLICITLY not-yet-built per founders), Mem=N (saved workflows ≠ agent-memory), Dep=N (cloud-first 'no desktop installs'; secondary macOS-desktop claims conflict, not relied on), Pack=N (builder + use-case-examples, no confirmed template library), Model=N (uses LLMs but no customer model-choice). Between Autotab (autonomous, 7.0) and Axiom (scripted RPA, 6.0): CopyCat = hybrid AI-RPA, young/narrow. Pricing: credits-based monthly; free trial (500 credits + proxies); Starter $100/mo (10k credits, proxy, 10 concurrent) — self_serve. Domain runcopycat.com. Score 4.5 (1F/7P/6N).

Capability coverage

4.5 / 14 capabilities · 32%

Integrations & Tool CallingCopyCat automates actions across any website including legacy portals and connects through webhooks and API triggers, real integration short of a documented broad library of native connectors, so partial.	Partial
Workflow OrchestrationCopyCat composes multi step browser workflows with branching, loops, and agentic steps that navigate dynamic sites intelligently, real orchestration short of fully autonomous end to end goal driven execution, so partial.	Partial
Knowledge Grounding & RAGCopyCat uses natural language models to interpret context and intent on dynamic web pages and extracts structured data, real grounding in page context short of a documented citation grounded retrieval system, so partial.	Partial
Human Oversight & GuardrailsCopyCat combines deterministic steps with agentic ones for reliability, but a runtime human approval, oversight, or guardrail enforcement engine could not be verified.	Unable to verify
Security, Identity & GovernanceCopyCat runs automations in the cloud with browser proxies, but named security certifications, role based access controls which the founders noted are not yet built, and identity governance could not be verified.	Unable to verify
Observability & AuditabilityCopyCat lets users test automations in the builder and watch them run in the cloud, real run visibility short of a documented full per action audit and logging system, so partial.	Partial
Memory & State PersistenceCopyCat saves reusable automation workflows, but a first class persistent agent memory could not be verified.	Unable to verify
Deployment & Data ResidencyCopyCat runs automations primarily in the cloud with no virtual machines or desktop installs, but a documented self host, on premises, or data residency deployment could not be verified.	Unable to verify
Prebuilt Agents, Templates & PacksCopyCat provides a visual builder and use case examples, but a library of prebuilt or cloneable automations could not be verified.	Unable to verify
Triggers & Channel CoverageCopyCat automations run on a schedule or are triggered by API and webhooks, real triggering short of broad event driven or customer facing channel coverage, so partial.	Partial
Model Flexibility & RoutingCopyCat uses natural language and language generation models to power its agentic steps, but user facing multi provider model selection or routing could not be verified.	Unable to verify
APIs, SDKs & MCP ExtensibilityCopyCat automations can be triggered via API and webhooks, a real extensibility surface short of a documented software development kit or Model Context Protocol server, so partial.	Partial
Testing, Debugging & OptimizationCopyCat lets users design, test, and refine automations in the builder and combine deterministic steps for reliability, real iterative testing short of a documented simulation or automated evaluation framework, so partial.	Partial
Browser & Computer UseCopyCat's browser agents navigate websites intelligently, handling dynamic pages where elements change, combined with precise deterministic actions, so full.	Full

Recent platform changes

No recent material changes tracked yet.

View all changes for CopyCat AI →

Pricing

Free trial (500 credits); paid from $100/mo

Public — exactHigh variable cost

Verified 2026-07-01

Data confidence: high