ANP

Executive Summary

ANP (Agent Network Protocol) is the most ambitious project in the agent coordination space — it doesn't just want to connect agents, it wants to rebuild the internet for an AI-native era. The three-layer architecture (identity, meta-protocol, application) goes deeper than any competing protocol, addressing not just communication but identity, negotiation, and semantic interoperability. ^[1]

The vision is compelling. The execution is early. ANP is a specification-first project with limited implementations, competing against Google's A2A which already has 150+ organizations and production SDKs. The project is alive but quiet: the main repo had commits as recently as June 10, 2026, and sits at ~1,300 stars — up only about 100 since February 2026. ^[1] The most notable development since then is on the standards track — a technical white paper on arXiv and an active role in the W3C AI Agent Protocol Community Group. ^{[2] [3]}

---

Product Overview

ANP defines a three-layer protocol stack for agent communication: ^[4]

Three-Layer Architecture

Key Differentiators

---

Technical Architecture

---

Strengths

• Deepest architecture — Three-layer design addresses identity, negotiation, and semantics systematically
• Decentralized identity — W3C DID avoids centralized auth bottlenecks and single points of failure
• Meta-protocol innovation — Agents negotiating their own communication format is genuinely novel
• Semantic discovery — Machine-readable capability matching goes beyond static Agent Cards
• Long-term vision — If the agent internet materializes, ANP's architecture is better suited than simpler protocols
• No vendor lock-in — Community-driven, no single corporate sponsor

---

Cautions

• Extremely early — Draft specification; the AgentConnect SDK is the main reference implementation ^[5]
• Complexity — W3C DID + meta-protocol + semantic web is a steep learning curve
• Small community — ~1,300 stars and ~90 forks as of June 2026, primarily China-based contributors ^[1]
• No enterprise adoption — No named users or production deployments as of June 2026
• Slow traction — Star growth of roughly 100 over four months (Feb–June 2026) signals limited international pickup, even as commits continue ^[1]
• Competing against momentum — A2A's 150+ organizations and 22K stars create strong network effects
• Implementation gap — Beautiful specs without battle-tested code
• Crypto-adjacent concerns — Explicitly disclaims cryptocurrency, but decentralized identity often attracts blockchain speculation
• No SDK maturity — No production-grade SDKs comparable to A2A's five-language offering

---

What Developers Say

No substantive first-person developer commentary on ANP could be found on Hacker News, Reddit, or similar English-language forums as of June 2026 — itself a signal of how niche the project remains outside China, where discussion happens mostly on domestic platforms like CSDN. The closest thing to independent practitioner assessment comes from third-party protocol surveys: an academic survey of agent interoperability protocols (Ehtesham et al.) treats ANP as one of the four major protocols alongside MCP, ACP, and A2A, noting that "ANP supports open network agent discovery and secure collaboration using W3C decentralized identifiers DIDs and JSON-LD graphs" — a positioning statement, not evidence of production use. ^[6]

---

Pricing & Licensing

---

Competitive Positioning

---

Ideal Customer Profile

Best fit:

• Researchers exploring decentralized agent networks
• Teams building permissionless agent ecosystems
• Organizations wanting vendor-neutral agent identity
• Long-term infrastructure projects with multi-year timelines

Poor fit:

• Anyone needing production agent coordination today
• Enterprise teams wanting proven, supported protocols
• Teams without distributed systems and semantic web expertise
• Projects with near-term delivery deadlines

---

Viability Assessment

---

Bottom Line

ANP has the most intellectually rigorous architecture of any agent coordination protocol. The three-layer design — decentralized identity, meta-protocol negotiation, semantic discovery — solves problems that A2A and ACP haven't even addressed yet. But architecture doesn't ship products.

Recommended for: Researchers and long-term infrastructure thinkers who want to understand where agent networking is headed.

Not recommended for: Anyone building production agent systems in the next 12 months.

Outlook: ANP's ideas will likely influence the space even if ANP itself doesn't win. Decentralized identity and meta-protocol negotiation are too important to ignore. The W3C Community Group route may be where that influence lands — ANP's concepts feeding a vendor-neutral standard rather than ANP-the-implementation winning adoption. ^[3]

---

Research by Ry Walker Research • methodology