Minicor
Startup that automates legacy Windows desktop software using AI agents.
On the record
Checkpoint: Microsoft's expected usage-priced, audit-logged agent SKU for regulated verticals
The call
Minicor is racing to lock in regulated customers before Microsoft's native Windows agents and frontier AI labs close the reliability gap it depends on.
Minicor is a software tool that watches your screen and does repetitive computer work for you, filling the gap between buggy AI agents (which fail a lot) and expensive corporate automation tools. The problem is that Microsoft already shipped nearly identical tools built directly into Windows in mid-2026, and companies will probably just use what comes with their operating system instead of buying Minicor separately. Minicor's only real chance is to lock in hospitals, car dealerships, or other heavily regulated businesses by offering detailed audit trails and compliance records that prove every action taken, before AI agents get reliable enough that even Microsoft's version works well enough for those industries too.
Why
01
AI pressure: OpenAI Computer-Using Agent, Anthropic Computer Use, Google Gemini Computer Use. Frontier agents click through Windows apps directly and keep closing the reliability gap Minicor is built to fill. Could AI do the core job? Yes, likely.
02
Bigger players: Microsoft (Windows 365 for Agents, Power Automate for Desktop). Microsoft bundled its own desktop-agent automation into Windows 11 Pro, the same job Minicor sells, on the OS Minicor depends on. Could a giant copy it? Yes, easily.
03
What protects it. Deep, certified footholds in regulated verticals like healthcare or auto dealer systems that a generic OS-level tool has not built yet.
What would change this call
Microsoft ships a usage-priced, audit-logged desktop-agent SKU for regulated industries by mid-2027. Would close Minicor's only viable compliance-vertical escape route.
Minicor names 2-3 healthcare or auto-dealer customers with SOC 2 or HIPAA certification in progress. Would show real vertical lock-in ahead of Microsoft's bundled offering.
OSWorld or similar benchmarks show frontier agents (OpenAI, Anthropic, Google) crossing 85-90% reliability on harder task sets. Would show the reliability gap Minicor depends on has closed.
Minicor names 2-3 healthcare or auto-dealer customers with SOC 2 or HIPAA certification in progress. Would show real vertical lock-in ahead of Microsoft's bundled offering.
OSWorld or similar benchmarks show frontier agents (OpenAI, Anthropic, Google) crossing 85-90% reliability on harder task sets. Would show the reliability gap Minicor depends on has closed.
Key numbers
- Raised $500K total (seed)
- Team of 7 to 9 people
- Claims 25,000 patients/day at one customer
Named risks
- Microsoft owns the Windows OS substrate and shipped a directly competing product (Windows 365 for Agents, Power Automate for Desktop) in May to June 2026.
- No patents, certifications, or exclusive data rights found in any source as of July 23 2026.
- Frontier computer-use models closing the reliability gap Minicor's product depends on, per OSWorld benchmark trends through July 2026.
- No named enterprise customers or independently verified traction metrics as of July 23 2026.
- Very small team of 7 to 9 people and only $500K raised, limiting runway against well-funded incumbents.
Rivals, as read in July 2026
| Rival | Direction | Evidence |
|---|---|---|
| Microsoft | losing to | Shipped Windows 365 for Agents (Jun 1 2026) and Power Automate for Desktop in Windows 11 Pro (May 26 2026), same problem, native to OS |
| OpenAI (Computer-Using Agent / ChatGPT agent mode) | losing to | Frontier OSWorld task completion rose from 34.5pct to 85.4pct over 14 months per Jul 8 2026 report |
| Anthropic (Computer Use) | contesting | Sep 4 2025 arXiv study found UiPath RPA still beats Anthropic Computer Use Agent on speed and reliability |
| Google (Gemini 2.5 Computer Use / Project Mariner) | losing to | Named as frontier computer-use competitor encroaching on same automation space as of Jul 8 2026 |
| UiPath / Automation Anywhere | contesting | Named legacy RPA big three that agentic startups position against, per Aug 12 2024 TechTarget report |
| Orby AI | contesting | 2022-founded peer using large action models against legacy RPA, with named customer JLL, per Aug 12 2024 report |
Questions people ask
- Is Minicor profitable in 2026?
- No revenue or profitability figures are public. Minicor has raised only $500K (vendor-claimed, YC seed check) as of July 23, 2026, and has 7 to 9 employees, consistent with a pre-revenue-disclosure early stage startup, not a profitable one.
- Could AI replace Minicor?
- Possibly. Minicor's product is a reliability wrapper around frontier computer-use models. As of July 8, 2026, OSWorld benchmarks show frontier models jumping from 34.5% to 85.4% task completion in 14 months, closing the exact reliability gap Minicor sells.
- Who is Minicor's biggest competitor?
- Microsoft, which shipped Windows 365 for Agents (June 1, 2026) and bundled Power Automate for Desktop into Windows 11 Pro (May 26, 2026), targeting the identical legacy Windows desktop automation problem Minicor addresses, directly inside the OS Minicor automates.
- Who owns Minicor?
- Minicor is privately held and venture-backed, founded in 2024 by CEO Faizaan Chishtie and CTO Saheed Akinbile. It is a Y Combinator portfolio company (batch listed as P26/X26), headquartered in San Francisco, with no institutional round beyond the YC seed check disclosed as of July 2026.
- What does Minicor actually do?
- Minicor builds AI computer-use agents that automate legacy Windows desktop software lacking usable APIs, such as EHRs, ERPs, dealer management systems, and QuickBooks/Sage/Citrix environments, using self-healing agents and a reflection agent to verify actions, per its March 22, 2026 Launch YC post.
- Does Minicor have any real customers?
- Minicor claims (vendor-claimed, unaudited) one healthcare customer running its architecture in production at 25,000 patients per day, as of March 2, 2026. No named enterprise logos or customer counts appear in any independent third-party source.
Timeline
- 2026-07-23Aggregator lists Minicor headcount at 9 and total funding at $500K
- 2026-07-08OSWorld 2.0 benchmark release resets best-model computer-use task completion to about 20.6 percent
- 2026-06-14Microsoft publishes guide on connecting legacy desktop apps to Copilot Studio agents
- 2026-06-01Microsoft ships Windows 365 for Agents targeting legacy Windows UI automation
- 2026-05-26Microsoft bundles Power Automate for Desktop into Windows 11 Pro
- 2026-05-22Startuphub profile details CTO Saheed Akinbile's background at Morgan Stanley and Ford
- 2026-03-22Minicor publicly launches on Launch YC and Hacker News
- 2026-03-02Minicor product site (laminar.run) publishes vendor-claimed accuracy and usage metrics
- 2025-09-04arXiv study finds UiPath RPA still outperforms Anthropic Computer Use Agent on enterprise workflows
- 2024-08-12TechTarget names UiPath and Automation Anywhere as legacy RPA big three amid agentic RPA rise
Revision history
27 July 2026
First assessed. Protection under test. The text above is fixed as of this date and is not edited afterwards.
In the same band
Sources
- When to use Windows 365 for computer-using agents (CUA) vs. RPA (2026-06-01)
- Launch YC: Minicor - Windows desktop automations at scale (2026-03-22)
- Scalable Desktop Automations (2026-03-02)
- Launch YC: Minicor - Self-healing desktop automations that scale (2026-03-22)
- Minicor - $500K Raised - Reviews & Alternatives (2026-07-23)
- Saheed Akinbile - Founder (2026-05-22)
- Computer-Using Agent
- AI Computer Use Benchmarks 2026: Top Agents Ranked (2026-07-08)
- Are LLM Agents the New RPA? A Comparative Study with RPA Across Enterprise Workflows (2025-09-04)
- Startup aims to upend old-school RPA with large action model (2024-08-12)
- Jobs at Minicor