Ephemeral Empire

ASC :: The coordination layer between orgs for the agent economy — one integration to connect, route, bill and govern any agent provider.

[Ephemeral Empire @MattLoseke]

Three important questions when thinking about securing dev machines:
1. What are devs doing?
2. What can dev machines do?
3. How are you monitoring actions taken?

[Zack Korman @ZackKorman]

People have messaged me saying "I like your videos but I don't understand what you're talking about."

So I took a step back and made a video showing how I write code and what that means for securing developer machines. If you feel lost, this is for you.

https://t.co/dBrfoI62Si

[Ephemeral Empire @MattLoseke]

It's agents all the way down: https://t.co/GCw5p23UnL

[Ephemeral Empire @MattLoseke]

<generated-reply> Valid point on HD seed fragility — key rotation without identity loss is a real problem worth solving. For now we handle it at the application layer (revoke + re-register), but the identity system is designed to be swappable.

The architectural bet we're making: coordination layer stays chain-agnostic and identity-layer-agnostic. If smart contract wallets win for agent identity, ASC can adopt them as an auth adapter. Same way settlement adapters let providers choose Lightning or Stripe.

Locking the coordination layer to any single chain's identity system is the one thing we're actively avoiding.

[Ephemeral Empire @MattLoseke]

@thedarshanjoshi building something around this: https://t.co/GCw5p23UnL
easy on-board via mcp

[Ephemeral Empire @MattLoseke]

@rldfst90 Built this into the coordination layer. Every agent call gets tracing — latency, retries, circuit breaker state, SLA compliance. Quality gates block settlement if delivery doesn't meet thresholds. open source: https://t.co/GCw5p23UnL

[Ephemeral Empire @MattLoseke]

<generated-reply>
Both. Agents self-declare capabilities at registration (endpoints, pricing, SLA terms) and those go into a queryable registry. Identity is secp256k1 keypairs with BIP-32 HD derivation — agents sign requests, counterparties verify.

The registry is the discovery layer, the crypto identity is the trust layer. Consumers query the registry to find agents, then verify identity before contracting. SLA +
quality gates enforce delivery before settlement clears.

Key difference from on-chain approaches: we're settlement-agnostic. Adapter pattern — Stripe for USD, Lightning via phoenixd for sats, more coming. The coordination layer shouldn't force a payment rail.

[Ephemeral Empire @MattLoseke]

This is the distribution layer for agent tooling and, these tools need to coordinate with each other: discovery, identity, billing, SLA enforcement across providers.

That's the layer we're building. Open source, MCP-native, 44 tools ready to plug in. https://t.co/GCw5p23UnL

[Claude @claudeai]

Introducing the Claude Marketplace, a way for enterprises to simplify their procurement of AI tools.

Now in limited preview.

[Ephemeral Empire @MattLoseke]

@claudeai Huge unlock for the agent ecosystem. Distribution was the missing piece for Claude-native tooling.

We ship an MCP server with 44 coordination tools — registry, billing, settlement, identity. Joining the partner waitlist now. https://t.co/GCw5p23UnL

[Ephemeral Empire @MattLoseke]

@iamnickelsteel1 I have a take on this: https://t.co/GCw5p23UnL Registry, coordination, identity, settlement — open source. The key design decision: the coordination layer has to be a trust broker, not a trust gatekeeper, otherwise agents just route around it. Early, looking for feedback

[Ephemeral Empire @MattLoseke]

@htahir111 Your three-layer decomposition maps cleanly to building cross-org agent coordination. ZenML solves the infra layer within an org - the unsolved piece is what happens when agent A in org 1 needs to contract agent B in org 2. Discovery, trust, settlement. Would be great to compare notes.

[Ephemeral Empire @MattLoseke]

@htahir111 "Infrastructure changes per org" — there's an emerging need for this layer. Been building it: registry, coordination, billing, identity, SLA enforcement across agents and orgs. Open source: https://t.co/GCw5p23UnL
easy mcp-based onboard, also have an sdk

[Ephemeral Empire @MattLoseke]

@naocfcloud @TechCrunch The "rogue agent" problem is a coordination problem. Agents go rogue when there's no layer governing discovery, contracting, and delivery verification.

Detection after the fact is one approach. Building the infrastructure so it can't happen is another. https://t.co/GCw5p23UnL

[Ephemeral Empire @MattLoseke]

@Luc_OneClaw Been building exactly this. Registry for discovery, coordination engine for multi-step handoffs, billing with escrow semantics, quality gates for delivery verification.

Open source, MCP server for onboarding: https://t.co/GCw5p23UnL

[Ephemeral Empire @MattLoseke]

#001 https://t.co/vhlazwrBEh

Gartner saw a 1,445% surge in multi-agent system inquiries from Q1 2024 to Q2 2025. Enterprises average 12 agents today, with 67% growth projected in two years. The coordination wall — the point where agent sprawl creates more complexity than value — hits at 5-10 agents within a single organization.

[Ephemeral Empire @MattLoseke]

@thedarshanjoshi Interesting...are these agents primarily inter-org? Does this approach work with them going intra-org as well?

[Ephemeral Empire @MattLoseke]

@iammichaelgrier @philvenables MCP and A2A handle connectivity but neither handle transitive trust across org boundaries. The coordination layer will emerge once protocols mature.

[Ephemeral Empire @MattLoseke]

MCP handles tool connectivity.
A2A handles agent communication.
ACP handles event-driven interaction.

None handles billing, SLAs, quality, or governance across organizational boundaries.

Three competing protocols. Still no coordination layer.