Pentagon
Pentagon is a workspace where teams of humans and AI agents get coordinated and tracked (YC P26)
On the record
Checkpoint: YC P26 Demo Day, where Pentagon's traction versus peers like Lyzr and Omnara becomes visible
The call
Pentagon sits on a busy, real market for agent coordination with a liked interface but no data, integration, or regulatory wall stopping Google or Microsoft from building the same thing.
Pentagon is a tool that helps businesses coordinate and manage multiple AI agents working together, kind of like a traffic controller for different AI tools. The problem is that Microsoft (which makes Copilot), Google (which has Agent Designer), and IBM (which has watsonx) can all add this coordination feature directly into their own products, which billions of people already use, so they don't need Pentagon. What could change things is if Pentagon gets customers locked in by making their specific workflows so dependent on Pentagon that switching away becomes painful, but Microsoft and Google can copy that playbook faster than Pentagon can build it.
Why
01
AI pressure: Frontier labs (OpenAI, Anthropic, Google) building native multi-agent handoff into their own agent SDKs. If OpenAI or Anthropic ship agent-to-agent coordination inside their SDKs, Pentagon's core feature becomes a free default. Could AI do the core job? Yes, likely.
02
Bigger players: Google (Agent Designer) and Microsoft (AutoGen, Copilot Studio). Google's Agent Designer already runs inside GenAI.mil; Microsoft's AutoGen and Copilot Studio cover similar agent-orchestration ground. Could a giant copy it? Yes, easily.
03
What protects it. A visual, spatial canvas for organizing agent teams that early users like, but nothing yet stops a bigger platform from copying it.
What would change this call
Microsoft or Google ships a native agent-org-workspace feature in Copilot Studio or Agent Designer within 6-12 months. Would show incumbents closing Pentagon's UX gap and eliminating its differentiation window.
Pentagon announces 3+ named enterprise customers with over $50K ARR within 12 months. Would show Pentagon achieving workflow lock-in strong enough to resist incumbent replication.
Lyzr AI publishes named enterprise customers or case studies in regulated verticals (finance, healthcare, legal). Would show Lyzr capturing the governance-focused market Pentagon needs to survive on.
Pentagon announces 3+ named enterprise customers with over $50K ARR within 12 months. Would show Pentagon achieving workflow lock-in strong enough to resist incumbent replication.
Lyzr AI publishes named enterprise customers or case studies in regulated verticals (finance, healthcare, legal). Would show Lyzr capturing the governance-focused market Pentagon needs to survive on.
Key numbers
- 137 of 196 YC P26 startups build AI agents
- Lyzr shipped rival Agent Control Plane
- Pentagon self-reports 8500+ persistent agents
Named risks
- No proprietary data, integrations, or shipped sandboxing to defend against platform giants.
- No revenue, customer logos, or third-party verified traction found as of mid-2026.
- Three near-identical domains (pentagon.run, joinpentagon.com, pentagonrun.com) with conflicting version numbers create brand confusion.
- Batch saturation: 137 of 196 YC P26 peers building competing agent products.
- Solo founder with a prior CEO resignation from marginfi in April 2024.
Rivals, as read in July 2026
| Rival | Direction | Evidence |
|---|---|---|
| Lyzr AI | gaining on | Shipped enterprise Agent Control Plane targeting banking, insurance, healthcare, 2026-07-10 |
| Google (Agent Designer) | losing to | Used to build 100,000+ agents inside a separate government platform as of 2026-07-07 |
| Microsoft (AutoGen) | contesting | Already shipping native agent orchestration into Copilot ecosystem as of 2026-07-09 |
| Omnara (YC peer) | contesting | Open-source, model-agnostic control plane for agents, same YC control-plane niche |
| IBM watsonx Orchestrate | contesting | Publishing agent orchestration platform content signaling platform ambitions, 2026-07-09 |
Questions people ask
- Is Pentagon (the YC startup) profitable in 2026?
- No revenue figures have been disclosed. Pentagon is a pre-launch stage Y Combinator P26 company that publicly launched via YC Launch on 2026-05-20 and offers a free Pentagon Studio download alongside a custom-quote enterprise tier, with no confirmed customers or revenue as of mid-2026.
- Who is Pentagon's biggest competitor?
- Lyzr AI shipped a near-identical 'Agent Control Plane' for enterprise agent governance on 2026-07-10, directly targeting the same buyers. Platform giants Microsoft (AutoGen), Google (Agent Designer), and IBM (watsonx Orchestrate) also already ship agent-orchestration tooling that overlaps with Pentagon's positioning.
- Could AI or a bigger platform replace Pentagon's product?
- Yes, this is a stated structural risk: frontier labs and cloud platforms (OpenAI, Anthropic, Google, Microsoft) are shipping native multi-agent coordination UX directly into their own agent SDKs, and Google's Agent Designer was already used to build 100,000+ agents inside a separate government platform as of 2026-07-07.
- Who founded Pentagon, the agent coordination startup?
- Edgar Pavlovsky, who previously founded the Solana DeFi protocol marginfi and resigned as its CEO on 2024-04-11 amid internal disagreements, per Decrypt. He also worked at Uber and Goldman Sachs before founding Pentagon, per Pentagon's YC profile.
- How many users or agents does Pentagon actually have?
- Pentagon's own marketing site (joinpentagon.com, snapshot dated 2026-03-02) claims 8,500+ Persistent Agents, 3,500+ Autonomous Agent Conversations, and 8,000+ Tasks Completed. These are vendor-claimed usage counters from Pentagon itself, not independently verified revenue or customer counts.
- Is Pentagon backed by Y Combinator?
- Yes, Pentagon is a member of Y Combinator's Spring P26 2026 batch and launched publicly through the Launch YC channel on 2026-05-20, per its official YC company and launch pages.
Timeline
- 2026-07-10Competitor Lyzr AI launches enterprise-focused Agent Control Plane
- 2026-06-16Report finds 137 of 196 YC P26 companies are building LLM agents
- 2026-05-20Pentagon publicly launches via Launch YC channel
- 2026-03-02Pentagon site (joinpentagon.com) claims 8,500+ persistent agents, v1.8.2
- 2024-04-11Founder Edgar Pavlovsky resigns as marginfi CEO amid disagreements
Revision history
26 July 2026
First assessed. Little protection left. The text above is fixed as of this date and is not edited afterwards.
In the same band
Sources
- Pentagon AI Hits 1.7M Users, Pilots Compliance Agents (2026-07-07)
- Lyzr Launches Agent Control Plane to Move Enterprise AI Agents From Pilot to Production (2026-07-10)
- YC P26 Batch: The Full Breakdown (2026-06-16)
- Launch YC: Pentagon - A coordination layer for humans and agents (2026-05-20)
- Introduction
- Pentagon (2026-03-02)
- Pentagon (2026-03-02)
- Pentagon: The control plane for agent-native work
- Solana Project Marginfi Withdrawals Top $214 Million After CEO's Resignation (2024-04-11)
- Omnara: The Control Plane for Agents
- AI Assistant Startups funded by Y Combinator (YC) 2026 (2026-07-25)
- Karpathy Hasn't Written Code Since December, He Just Directs AI Agents Now (2026-03-23)