Best of · Developer & infrastructure
Best desktop and mobile app automation for AI agents
All 6 ranked desktop and mobile app automation on the Anchor benchmark, with a pick for each need and where each one falls short. Scores come from public evidence, re-checked as vendors change.
- 6 ranked
- 1 agent-ready
- 2 hosted endpoints
- Updated 14 October 2026
Top three
Picks by need
Worked out from the scores, prices and facts, so they change when the research does.
The shortlist
| # | Tool | Grade | Best for | Price | Where |
|---|---|---|---|---|---|
| 1 | Robot Framework Robot Framework Foundation |
BB 76.4 | Agents that write or run keyword-driven test and RPA suites across web, desktop and mobile through libraries. | Free · OSS | Library |
| 2 | UiPath UiPath |
B 68.2 | An agent that needs to drive Windows desktop or legacy applications through robots a company already runs, started as MCP tools or API jobs. | $25 / mo | Hosted |
| 3 | Robocorp Sema4.ai |
B 65.5 | Python RPA on Windows desktops, browsers and documents, written or run by a coding agent. | Free · OSS | Local |
| 4 | AutoHotkey AutoHotkey Foundation |
D 52.8 | An agent automating Windows desktop apps on one machine by writing short scripts. | Free · OSS | Local |
| 5 | Automation Anywhere Automation Anywhere |
D 50.9 | A company already running Automation 360 bots that wants an assistant such as Copilot or Claude to find and start them. | Freemium | Hosted |
| 6 | AutoIt AutoIt Consulting |
D 46.3 | An agent automating Windows desktop apps on one machine, and AutoItX lets C# or PowerShell code call the same functions. | Free | Local |
How to choose
- Which platforms and appsCheck which operating systems, application types and mobile platforms are supported, since a tool limited to Windows cannot reach an app on macOS or a phone.
- How an agent starts a runCheck whether runs can be started and read through an API, CLI or MCP server, because a tool driven only from its own studio cannot be called by an agent.
- Credentials and unattended runsCheck where the logins a robot uses are stored and who can read them, since an unattended robot can reach everything its account can.
- Licence and runtime costCheck whether robots are licensed per machine, per user or per run, since unattended automation is often priced separately from building it.
How the benchmark tests this category. In this run, public evidence against the published checklist. When the task suites run, the same form-filling task in a desktop application started by an agent through each listing's API, CLI or MCP server. We check whether the run completes, how failures are reported, where credentials are kept and what a run costs.
Each one in detail
Robot Framework
BB 76.4/100Robot Framework is an open-source automation framework for test automation and robotic process automation, run from the robot command line or Python. Its keyword libraries reach browsers, desktop applications and mobile apps.
Verdict A stable 7.5 release, passing unit and acceptance tests, JSON schemas for results and library docs, and an llms.txt make it a dependable base for agent-written automation. Desktop and mobile control come from separate community libraries, there is no official MCP server, and no security policy or security.txt was found.
Best for Agents that write or run keyword-driven test and RPA suites across web, desktop and mobile through libraries.
Strengths
- Version 7.5 of 14 September 2026, marked Production/Stable on PyPI, with Python 3.8 to 3.15 listed
- Unit and acceptance test workflows on CPython and PyPy passed on master on 6 October 2026
- JSON schemas for run results, executable suites and Libdoc library specs are published in doc/schema
Weaknesses
- Desktop and mobile control depend on separate libraries, such as AppiumLibrary and AutoItLibrary, maintained outside the core project
- No official MCP server. MCP servers for Robot Framework found on PyPI and Glama are community projects
- No SECURITY.md, no security.txt on robotframework.org and no published repository advisories
Price Free · OSSAuth Nonex402 NoLibrary
UiPath
B 68.2/100UiPath is a robotic process automation platform whose hosted Orchestrator runs robots that operate desktop, web and Windows applications. Outside agents start and monitor jobs through the Orchestrator API or folder-scoped MCP servers in Automation Cloud.
Verdict Orchestrator exposes published RPA workflows, agents and API workflows as MCP tools over Streamable HTTP, with MCP OAuth, scoped PATs and an OpenAPI 3.0 API. Prices above the $25 Basic plan need a sales conversation, and the status page logged several multi-hour incidents between 10 September and 13 October 2026.
Best for An agent that needs to drive Windows desktop or legacy applications through robots a company already runs, started as MCP tools or API jobs.
Strengths
- Folder-scoped MCP servers turn chosen processes, agents and API workflows into tools at a URL of the form cloud.uipath.com/{org}/{tenant}/agenthub_/mcp/{folderKey}/{slug}
- MCP OAuth flow with Client ID Metadata Documents, dynamic registration and RFC 8707 resource binding, plus PATs scoped to OR.Jobs and OR.Execution
- Rate limits published per endpoint, with Retry-After and X-RateLimit-Remaining headers on 429 responses
Weaknesses
- Standard and Enterprise prices are not published, and robot licences are sold through sales
- Status history shows robot job failures in Europe for over seven hours on 10 September 2026 and a certificate expiry hitting several services on 13 October 2026
- No idempotency key was found for StartJobs, and the MCP pages say nothing about readOnlyHint or destructiveHint
Price $25 / moAuth OAuthx402 NoHosted
Robocorp
B 65.5/100Robocorp's open-source automation stack for Python, now run by Sema4.ai, with libraries for Windows desktop, cross-platform desktop, browser and document automation. Tasks run locally through rcc or Python, with Control Room as the hosted orchestrator.
Verdict Apache-2.0 libraries with passing CI, releases in the last month and an agent skill in the repository make it practical for agent-written desktop automation. robocorp.com now redirects to Sema4.ai, whose docs, pricing, terms and rcc downloads sit behind a bot check, and recent library issues mostly go unanswered.
Best for Python RPA on Windows desktops, browsers and documents, written or run by a coding agent.
Strengths
- Apache-2.0 packages on PyPI, rpaframework 33.0.2 (4 October 2026) and robocorp 3.1.4 (28 September 2026)
- RPA.Windows and robocorp.windows drive Windows controls by locator, and RPA.Desktop is a cross-platform library for keyboard, mouse and screen
- rpaframework's main and Windows CI workflows passed on master on 4 October 2026
Weaknesses
- robocorp.com redirects to sema4.ai, which answered our reader with a bot check, so the docs, Control Room pricing and terms were not read
- The repository's own agent skill says the rcc release page blocks automated requests and asks the user to fetch the download link
- Of the 15 newest open rpaframework issues and pull requests, 8 had no reply, including Windows automation issues open since 2024
Price Free · OSSAuth Nonex402 NoLocal
AutoHotkey
D 52.8/100AutoHotkey is a free, open-source scripting language and interpreter for Windows that sends keystrokes and mouse input, manages windows and controls, and defines hotkeys. Agents run scripts through the AutoHotkey64.exe command line, including from standard input.
Verdict An agent can write a script and run it through the command line, with a script read from standard input, /Validate and /ErrorStdOut for clean failures, and a function reference of 356 pages. It runs on Windows only with the user's full rights, and the repository has no CI or test suite.
Best for An agent automating Windows desktop apps on one machine by writing short scripts.
Strengths
- Scripts can be piped to the interpreter on standard input with
*as the filename /Validatechecks a script without running it and exits non-zero on a load error- Four 2.0.x bug-fix releases between 28 August and 9 October 2026, each in a dated changelog
Weaknesses
- Windows only, and scripts act with every right of the user who runs them
- No CI workflow or test suite in the repository
- No security policy, security.txt or published advisory was found
Price Free · OSSAuth Nonex402 NoLocal
Automation Anywhere
D 50.9/100Automation Anywhere's Automation 360 is a cloud robotic process automation platform whose Control Room deploys bots that operate desktop and web applications. Outside agents run automations through the Control Room REST API or its MCP inbound endpoint.
Verdict The Control Room's /mcp endpoint gives an MCP client three standard tools to discover, run and read automations under RBAC, with each run in the audit log. No API rate limits, error catalogue, SLA or prices were found in public pages, and DiscoverAutomation and RunAutomation need an Enterprise licence with the Process Reasoning Engine in the US or Europe.
Best for A company already running Automation 360 bots that wants an assistant such as Copilot or Claude to find and start them.
Strengths
- MCP inbound at https://<control-room>/mcp with DiscoverAutomation, RunAutomation, GetAutomationResult and custom inbound tools with typed input and output variables
- Each run started through MCP is logged as device acquired, bot sent, deployed and finished in the Control Room audit log
- Written API deprecation policy, with at least two years of support and at least a year's notice before end of life
Weaknesses
- No API rate limits, 429 handling or retry guidance found in the public Control Room API pages
- The published response code table lists only generic HTTP statuses, with no error bodies
- DiscoverAutomation and RunAutomation need an Enterprise licence with the Process Reasoning Engine, supported in the US and Europe only
Price FreemiumAuth OAuth or keyx402 NoHosted
AutoIt
D 46.3/100AutoIt v3 is a freeware BASIC-like scripting language for automating the Windows GUI with simulated keystrokes, mouse movement and window and control manipulation. Agents run scripts through AutoIt3.exe, and AutoItX exposes the same functions through a DLL and COM control.
Verdict An agent can run AutoIt scripts or single lines from the command line, or call AutoItX from C# or PowerShell, with each function's return values and @error codes documented. The source is closed, the last release was 7 September 2025, and scripts act with the full rights of the Windows user.
Best for An agent automating Windows desktop apps on one machine, and AutoItX lets C# or PowerShell code call the same functions.
Strengths
/AutoIt3ExecuteLineruns one line of code without a script file- AutoItX exposes the functions through a DLL and COM control with a C# assembly and PowerShell cmdlets
- Each function page documents return values and
@errorcodes, with an example
Weaknesses
- Closed source under a licence that forbids reverse engineering
- Last release 3.3.18.0 on 7 September 2025, three years after 3.3.16.1
- Windows only, and scripts act with every right of the user who runs them
Price FreeAuth Nonex402 NoLocal
Head to head
- Robot Framework vs UiPath BB 76.4 vs B 68.2
- Robocorp vs Robot Framework B 65.5 vs BB 76.4
- AutoHotkey vs Robot Framework D 52.8 vs BB 76.4
- Automation Anywhere vs Robot Framework D 50.9 vs BB 76.4
- Robocorp vs UiPath B 65.5 vs B 68.2
- AutoHotkey vs UiPath D 52.8 vs B 68.2
- Automation Anywhere vs UiPath D 50.9 vs B 68.2
- AutoHotkey vs Robocorp D 52.8 vs B 65.5
- Automation Anywhere vs Robocorp D 50.9 vs B 65.5
- AutoHotkey vs Automation Anywhere D 52.8 vs D 50.9
Questions
What are the highest-rated desktop and mobile app automation for AI agents?
Robot Framework has the highest benchmark score of the 6 ranked desktop and mobile app automation, 76.4 (BB). UiPath is second with 68.2 (B).
How many desktop and mobile app automation are agent-ready?
1 of the 6 ranked here grade BB or better, the bar for agent-ready on the Anchor benchmark.
Which desktop and mobile app automation accept x402 payments?
None of the ranked listings here accepts x402 for its main call yet.
How is this list ranked?
By the Anchor benchmark score out of 100, a weighted mean of the scored categories minus deductions for negative events, from public evidence re-checked as vendors change. Listings cannot pay for a place. The latest assessment behind this page is from 14 October 2026.
How this list is made
The order is the Anchor benchmark score, the same number as on each listing and in the top list. Each listing is graded from public evidence against the benchmark checklist, and the picks above are worked out from those grades, prices and facts. No listing pays for its place, and paid audits or listing help never change a score.
Full ranked table · 15 head-to-head comparisons · Best tools in every category