We build the rails the
agent economy will run on.
DCS AI Technologies is a Dubai-registered company building the open infrastructure layer for AI agents — provenance, identity, memory, treasury. The standard is free. The reference implementations are open. The business is the operational tier.
The company
DCS AI Technologies L.L.C was incorporated in Dubai, UAE on 21 May 2026 under Commercial License No. 1624450 issued by the Dubai Department of Economy & Tourism. We operate two public brands. DCS AI (this domain, dcsai.ai) is the corporate face — investor materials, press, the parent narrative. DCS Labs (dcslabs.ai) is the builder-facing surface — the R+2 standard, the MCP server, the mint funnel, live stats. Same company, two surfaces, different audiences.
We launched publicly on May 21, 2026, after a ~41-day intensive solo, AI-assisted build. Read the full launch story at dcsai.ai/blog/coming-live.
The bet
AI agents are about to do real economic work. They'll hold budgets, sign contracts, negotiate with each other, remember what they did six months ago, and present audit trails when regulators ask. None of that infrastructure exists today. That's the gap DCS was built to close.
Specifically, the bet is on four interlocking primitives:
- Provenance — every action signed, every action verifiable (R+2 Open Provenance Standard)
- Identity — non-transferable on-chain identity for each agent (TRDWorkerSBT on Base mainnet)
- Memory — cryptographically signed cross-session memory (Sovereign Memory)
- Treasury — agent-to-agent settlement layer with a 2.5% fee on USDC, on Base
Each requires the others to be meaningful. Cloning any one of them doesn't get you the stack; cloning all four with cryptographic linkage is an 18-month project.
The founder
Deepak Dudi is the solo founder of DCS AI Technologies. Indian-origin, Dubai-registered. Built a prior AI infrastructure venture (operating system for vertical-AI workforces) before DCS. Reachable directly at [email protected].
"Eight weeks ago I stopped tweeting and started building. The point was simple — I didn't want to claim anything I couldn't prove. Every claim DCS makes today resolves to a Basescan transaction, an npm package version, or a published RFC-style specification. That's the constraint I set on myself in January. It's the constraint we'll hold to."
Operating principles
1. Open standard, paid infrastructure
The R+2 specification is MIT-licenced and free to adopt. We don't make money from the standard. We make money from the reference infrastructure — Sovereign Memory, Agent Treasury, managed identity. Anyone is free to clone our infrastructure too; we'll compete on quality and reliability, not lock-in.
2. Verifiable, not asserted
Every metric we publish has a public URL behind it. Every claim has a transaction hash, a package version, or a published spec. We do not produce marketing-math screenshots.
3. No model training on user data
Memory writes, agent action data, and user content are never used to train AI models. This is a permanent commitment, not a current policy that we might change.
4. Sovereign-deployment friendly
The full stack is air-gappable. Governments, defence agencies, and regulated industries that can't allow outbound connectivity can run DCS on-premise with no functional degradation.
5. Founder-led, self-funded, until product-market fit
DCS will stay solo and self-funded until product-market fit is clear. Hires are paced against founder capital plus pilot revenue — no fundraising dependency. First 2-3 senior hires (memory, cryptography, on-chain) will close as traction warrants. Sales runs through the founder until we have a repeatable motion to hand off.
Where to reach us
Builders: [email protected]
Press: [email protected]
Investors (inbound only — not raising): [email protected]
Standards bodies + governments: [email protected]
Cold inquiries: [email protected]
Founder direct: [email protected]