DEV Community: Barbara Wu

The Billion-Dollar Blind Spot: Why Nobody Is Using AI to Accelerate Tooth Regeneration

Barbara Wu — Fri, 06 Mar 2026 05:01:42 +0000

The Billion-Dollar Blind Spot: Why Nobody Is Using AI to Accelerate Tooth Regeneration

The world's first drug to regrow human teeth is in clinical trials. AI has designed antibodies that are already in human testing. These two revolutions have never met. Here's why — and why that's about to change.

Teeth That Grow Back: It's Actually Happening

In October 2024, something remarkable began at Kyoto University Hospital: the world's first clinical trial of a drug designed to make humans grow new teeth.

The drug, TRG-035, is a monoclonal antibody that targets a protein called USAG-1 (gene name: SOSTDC1). USAG-1 acts as a molecular brake on tooth development — it simultaneously blocks both the BMP and Wnt signaling pathways, which are essential for tooth formation. By neutralizing USAG-1, the antibody releases dormant tooth buds that most humans carry but never develop, allowing new teeth to emerge.

The science behind this is elegant. In 2007, researchers led by Dr. Katsu Takahashi at Kyoto University showed through targeted gene knockout experiments that mice lacking USAG-1 grow supernumerary teeth — extra teeth that develop from rudimentary tooth germs that are normally suppressed. By 2021, they had published in Science Advances showing that a single injection of an anti-USAG-1 antibody could induce whole new teeth in mice, with the key insight that blocking USAG-1's interaction with BMP (rather than Wnt) was sufficient to trigger tooth regeneration.

The Phase I trial enrolled 30 men aged 30-64, each missing at least one molar, over an 11-month study period. Toregem BioPharma, the Kyoto University spinoff driving commercialization (a team of about 13 people), has partnered with WuXi Biologics for manufacturing and has already received FDA Pre-IND feedback for US clinical trials. In September 2025, TRG-035 received orphan drug designation from Japan's Ministry of Health for severe congenital tooth agenesis. The target: commercial availability by 2030.

Meanwhile, in China, a multicenter randomized controlled trial published in 2025 demonstrated that allogeneic dental pulp stem cell injections significantly improved periodontal bone defects (26.81% improvement vs. 17.43% for controls). At the Air Force Medical University, Professor Jin Yan's team achieved the world's first functional full-length dental pulp regeneration using stem cells from exfoliated deciduous teeth.

Tooth regeneration is no longer science fiction. It's in human trials.

The AI Drug Discovery Revolution: Proven and Accelerating

At the same time, AI-driven drug discovery has crossed a critical threshold — from theoretical promise to human clinical validation.

Absci has used its generative AI platform, Origin-1, to design full-length monoclonal antibodies entirely from scratch — no prior known binders needed. Two of these AI-designed antibodies are already in human clinical trials:

ABS-101 (anti-TL1A for inflammatory bowel disease): Phase 1, first participants dosed May 2025
ABS-201 (anti-PRLR for androgenetic alopecia): Phase 1/2a, first participants dosed December 2025

Insilico Medicine achieved another milestone: its AI-discovered drug Rentosertib (ISM001-055), a novel TNIK inhibitor for idiopathic pulmonary fibrosis, became the first drug with both target and molecule discovered by generative AI to enter Phase II clinical trials. The entire journey from target discovery to Phase 1 took under 30 months — roughly half the traditional timeline. Positive Phase IIa results were announced in late 2024.

The money follows the results: Generate Biomedicines signed a $1B+ deal with Novartis in September 2024. Isomorphic Labs (Alphabet's AI drug discovery spinoff) raised $600M in March 2025. The field is exploding.

As of early 2026, no AI-designed drug has yet received FDA approval. But the first is widely expected within 1-2 years.

The Gap: Two Revolutions That Have Never Met

Here is the striking fact at the center of this article:

There is not a single project — on GitHub, Kaggle, or Hugging Face — that applies AI drug discovery techniques to tooth regeneration.

I conducted a systematic search across all three platforms. Here's what I found:

GitHub: Only 3-4 repositories are even tangentially related to tooth regeneration biology:

Ruohola-Baker-lab/Tooth_sciRNAseq — Single-cell atlas of human tooth development (the most relevant, from a Developmental Cell 2023 paper that generated enamel proteins from iPSCs)
TheMoorLab/Tooth — Single-cell atlas of periodontal and dental pulp tissue
fberio/Genetic-building-of-teeth-and-odontodes — Gene expression database for vertebrate tooth development

Zero projects involving USAG-1/SOSTDC1 computational analysis, molecular docking, or AI-driven antibody design for tooth regeneration targets.

Kaggle: ~30 dental datasets found — all focused on X-ray image segmentation and cavity detection. Zero datasets related to tooth regeneration, dental stem cells, or USAG-1.

Hugging Face: Mature dental imaging models exist (OralGPT-Omni, Dental-GPT-OSS-20B, OralSeg). But for tooth regeneration? Nothing.

The pattern is clear:

Dental imaging AI:     ==================== Very mature
Dental diagnostic LLMs: ========             Emerging
Tooth regeneration AI:                       Empty

Why the Gap Exists

This isn't because people are lazy or unaware. There are real structural reasons:

1. The Discovery Didn't Need AI

The USAG-1 target was identified through classical developmental biology: gene knockout mice, phenotype observation, and hybridoma antibody screening. The five monoclonal antibodies (#12, #16, #37, #48, #57) were generated by immunizing USAG-1 knockout mice and fusing lymphocytes with SP2/0 myeloma cells — a technique from the 1970s. Most of this work happened before AlphaFold2 matured in 2020.

AI excels at optimization within known frameworks. It's less suited for the kind of serendipitous biological discovery that led to the USAG-1 breakthrough.

2. Data Desert

AI needs data. Tooth regeneration has almost none in machine-readable form:

No standardized datasets of tooth developmental gene expression
No large-scale screens of USAG-1 binding compounds
No public repositories of dental stem cell differentiation outcomes
The entire field produces perhaps a few dozen papers per year

Compare this to oncology, where millions of compounds, thousands of protein structures, and decades of clinical data are available for AI training.

3. Organ-Level Complexity

A tooth is not a single tissue. It's five distinct tissues — enamel, dentin, pulp, cementum, and periodontal ligament — each derived from different embryonic cell lineages (enamel from ectoderm, the rest from ectomesenchyme), requiring precise spatial organization and temporal coordination.

Bone regeneration = regenerating one tissue type. Skin regeneration = 2-3 layers. Tooth regeneration = coordinating five different tissues simultaneously. This approaches the complexity of whole-organ regeneration.

4. The Amelogenin Problem

The key protein for enamel formation — amelogenin, which comprises ~90% of the enamel matrix — is an intrinsically disordered protein (IDP). It has no fixed 3D structure. This is precisely AlphaFold's weakest area: the model cannot reliably predict disordered or dynamic substructures, and AlphaFold3 still shows ~22% "hallucination" rates for IDP regions. Since enamel regeneration depends on understanding amelogenin's self-assembly behavior, this creates a fundamental computational bottleneck.

5. Signaling Pathway Nightmare

Tooth development involves the interplay of four major signaling pathways — BMP, Wnt, FGF, and Shh — with extensive crosstalk. The same signal can be synergistic or antagonistic depending on cellular context, timing, and spatial location. This context-dependent combinatorial explosion is beyond current computational modeling capabilities.

6. Small Market, Low Priority

Dental implants already work well (>95% success rate over 5-10 years). Congenital tooth agenesis affects perhaps hundreds of thousands of people globally, compared to tens of millions for cancer. AI drug discovery companies like Absci and Insilico rationally prioritize oncology and autoimmune diseases — that's where the market is.

What AI Could Actually Do

Despite these challenges, there are clear opportunities where AI could meaningfully accelerate tooth regeneration — not by replacing wet-lab biology, but by augmenting it:

1. Better Antibodies Against USAG-1

TRG-035 was found through traditional hybridoma screening. It may not be optimal. Absci's Origin-1 platform has demonstrated the ability to design antibodies de novo against zero-prior epitopes. Applying similar AI antibody design to USAG-1 could yield:

Higher affinity antibodies (tighter binding = lower dose = fewer side effects)
Better selectivity (blocking BMP interaction without affecting Wnt, which is the key therapeutic mechanism)
Improved pharmacokinetics (longer half-life, better tissue distribution)

Important caveat: USAG-1 itself (unlike amelogenin) is a structured protein belonging to the sclerostin family with a cystine knot domain. Its AlphaFold-predicted structure is available. This means AI-driven antibody and small molecule design against USAG-1 is technically feasible.

2. Small Molecule Alternatives

An antibody drug like TRG-035 is expensive — perhaps thousands of dollars per treatment. If AI could design a small molecule inhibitor of USAG-1 (blocking its interaction with BMP), the cost could drop by 100x, transforming it from a niche orphan drug into a mass-market therapy.

Tools exist for this: AlphaFold 3 for protein-ligand complex modeling, generative chemistry models like DiffDock and RFdiffusion for candidate generation, and ADMET prediction models for drug-likeness screening. The pipeline is well-established in other therapeutic areas; it just hasn't been pointed at USAG-1.

Note: For small molecule docking against USAG-1, tools like AutoDock Vina are applicable. However, for modeling USAG-1's protein-protein interactions with BMP or LRP5/6, specialized tools like HADDOCK or PatchDock would be needed.

3. Mining the Dental Single-Cell Atlas

Rich single-cell transcriptomic data for human tooth development already exists (e.g., the Developmental Cell 2023 census). Foundation models like scGPT and GeneFormer — pre-trained on 33M+ cells — could be applied to:

Identify whether adult jawbone tissue retains dormant tooth progenitor cells (which would expand the addressable patient population from congenital agenesis to all adults with missing teeth)
Predict optimal reprogramming factor combinations to convert jawbone mesenchymal cells into tooth bud cells
Discover new regulatory targets in the BMP/Wnt/FGF/Shh signaling network

4. Accelerating Clinical Trials

AI could compress the Phase I-III timeline through:

Synthetic control arms using historical patient data (reducing required enrollment)
Adaptive trial design with real-time ML analysis of interim data
Digital twins predicting individual patient response
AI toxicology modeling potential side effects of USAG-1 inhibition on BMP/Wnt pathways (both are implicated in cancer — understanding the safety profile is critical)

What's Available Right Now

For anyone who wants to start working on this, here's the public data and tooling:

Data

Resource	What It Contains
AlphaFold DB: Q6X4U4	Predicted 3D structure of USAG-1/SOSTDC1
Murashima-Suginami et al., 2021	Antibody binding sites, functional classification, animal experiment data
Tooth_sciRNAseq (GitHub)	Human tooth development single-cell atlas + iPSC enamel organoid analysis
TheMoorLab/Tooth (GitHub)	Periodontal and dental pulp single-cell atlas
UniProt, KEGG, Reactome	BMP/Wnt pathway data (decades of accumulated knowledge)

Tools

Tool	Application
AlphaFold 3	Protein-ligand complex modeling for USAG-1
AutoDock Vina	Small molecule docking against USAG-1 binding pockets
HADDOCK / PatchDock	Protein-protein docking (USAG-1 ↔ BMP / LRP5/6)
DiffDock / RFdiffusion	Generative molecular design
scGPT / GeneFormer	Single-cell transcriptomic foundation models
ProteinMPNN / ProtGPT2	Protein sequence design
Scanpy / Seurat	Single-cell data analysis

The Competitive Landscape

Number of open-source projects combining AI drug discovery with tooth regeneration: zero.

The Opportunity

Let me be direct about what this is and isn't:

What this isn't: A guaranteed path to regrowing teeth faster. The biological challenges are real. Amelogenin is an IDP that AlphaFold can't model well. The signaling pathway complexity is daunting. Wet-lab validation is irreplaceable.

What this is: A clear case of cross-domain information asymmetry. The people who understand USAG-1 biology (developmental biologists in Kyoto, stem cell researchers in China) don't have AI drug discovery expertise. The people who build AI antibody design platforms (Absci, Isomorphic Labs) don't know this target exists. The first person or team to bridge this gap — even with basic computational analysis — will be doing something no one else in the world has done.

The most likely path to impact:

Use AlphaFold to model the USAG-1/BMP binding interface — understand where TRG-035 binds and how it could be improved
Apply generative models to design candidate small molecules against the same binding site — potentially replacing an expensive antibody with an affordable pill
Mine the dental single-cell atlas with foundation models — answer whether adults retain activatable tooth progenitor cells

Each of these is a publishable result. Together, they could meaningfully accelerate the timeline from 2030 to 2027-2028 — not by skipping biological validation, but by giving biologists better starting points.

We're Building This in the Open

I've started an open-source project to begin this work: Open Tooth Regeneration (link will be live soon).

The initial scope:

Curated dataset of all public USAG-1/SOSTDC1 structural and functional data
AlphaFold structure visualization and binding site analysis
Basic molecular docking experiments with AutoDock Vina
Integration of public dental single-cell transcriptomic data
A comprehensive literature review connecting tooth regeneration biology with AI capabilities

What I need: Collaborators with structural biology, medicinal chemistry, or dental developmental biology expertise. The AI tooling is ready. The data is public. What's missing is the interdisciplinary bridge.

If you work in any of these areas and find this interesting, open an issue or reach out. This is the kind of problem that no single discipline can solve alone — but the pieces are all there, waiting to be assembled.

All facts in this article have been individually verified against primary sources. See the fact-check document for detailed source attribution for every claim made above.

Sources:

Murashima-Suginami, A. et al. (2021). Anti-USAG-1 therapy for tooth regeneration through enhanced BMP signaling. Science Advances, 7(7), eabf1798. Link
Toregem BioPharma. Clinical trial initiation (October 18, 2024). Link
Toregem BioPharma. TRG-035 Orphan Drug Designation (September 29, 2025). Link
Toregem BioPharma. FDA Pre-IND Response (November 2025). Link
WuXi Biologics & Toregem BioPharma MOU (October 2022). Link
Absci. ABS-101 Phase 1 First Dosing (May 2025). Link
Absci. ABS-201 Phase 1/2a HEADLINE Trial (December 2025). Link
Insilico Medicine. First Generative AI Drug Enters Phase II. Link
AlphaFold Protein Structure Database: SOSTDC1 (Q6X4U4). Link
Pagella, P. et al. Single-cell atlas of healthy human teeth. Developmental Cell, 2023. Link
Applications of AI in regenerative dentistry. Frontiers in Cell and Developmental Biology, 2024. Link
EBI. AlphaFold: Strengths and Limitations. Link

MCP vs A2A: The Two Protocols Defining the AI Agent Economy

Barbara Wu — Fri, 06 Mar 2026 04:40:03 +0000

One lets agents "use tools," the other lets agents "talk to each other" — understand these two protocols, and you understand the foundation of the 2026 AI Agent ecosystem.

In March 2026, the hottest topic in the AI Agent space isn't which model is stronger — it's the protocol wars. Anthropic's MCP (Model Context Protocol) and Google's A2A (Agent2Agent Protocol) are defining the infrastructure of the Agent economy.

As a developer who has contributed code to LangChain and Haystack, and runs agents on OpenClaw daily, I want to decode from a practitioner's perspective: what problems do these two protocols solve? How do they work together? What do they mean for developers?

First, A Story: Why Do We Need Protocols?

Imagine you run a company with three AI Agent employees:

Agent R: Market research — good at search and data analysis
Agent C: Coding — can operate GitHub and databases
Agent F: Finance — can access ERP and banking systems

Problem 1: Agent R needs to query sales data from the database — how does she connect?
→ This is what MCP solves: letting agents use tools.

Problem 2: Agent R finishes her analysis and needs to hand results to Agent C to write an automation script — how do they communicate?
→ This is what A2A solves: letting agents talk to each other.

Without MCP, agents are "brains without hands." Without A2A, agents are "isolated islands." Together, they form a complete Agent economy.

MCP: Giving Agents Hands and Feet

What is MCP?

MCP (Model Context Protocol) is an open protocol launched by Anthropic in November 2024.

MCP is the USB standard of the Agent world — one protocol to connect all tools.

Before MCP, every AI app needed separate integration code for each tool. MCP unifies this: build one MCP Server, and any MCP-compatible agent can use it.

Architecture: Three-Layer Design

┌──────────────────────────────────┐
│          MCP Host                │  ← App running the agent
│  (Claude Desktop, VS Code)       │    (like the "OS")
├──────────────────────────────────┤
│         MCP Client               │  ← Communication middleware
│   (Protocol impl, JSON-RPC)      │    (like a "driver")
├──────────────────────────────────┤
│         MCP Server               │  ← Tool provider
│  (GitHub, Database, Slack...)    │    (like a "USB device")
└──────────────────────────────────┘

Key design: The Host acts as a Security Broker — all agent-to-resource interactions go through the Host. Like an OS controlling hardware access — agents can't bypass the Host to directly operate tools.

Three Core Primitives

Primitive	Purpose	Example
Tools	Functions agents can call	`search_web(query)`, `create_issue(title)`
Resources	Data agents can read	Database tables, file contents, API responses
Prompts	Pre-defined interaction templates	"Summarize the following data in a table..."

Communication: JSON-RPC 2.0

MCP uses JSON-RPC 2.0 under the hood — stateless, transport-agnostic, simple and reliable. A real tool call looks like:

// Agent → MCP Server (request)
{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "query_database",
    "arguments": {
      "sql": "SELECT product, SUM(revenue) FROM sales GROUP BY product"
    }
  }
}

// MCP Server → Agent (response)
{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "content": [{
      "type": "text",
      "text": "Product A: $120,000\nProduct B: $89,000\nProduct C: $210,000"
    }]
  }
}

November 2025 Major Update

Tasks primitive: Async long-running operations (previously sync-only)
OAuth 2.1: Enterprise-grade authentication
Resource Indicators: Prevents malicious servers from stealing access tokens

Ecosystem Status

SDK downloads: 97M+/month (Python + TypeScript)
Official MCP Servers: 50+
Supported Hosts: Claude Desktop, VS Code, Cursor, Windsurf, OpenClaw...

A2A: Letting Agents Talk to Each Other

What is A2A?

A2A (Agent2Agent Protocol) was launched by Google in April 2025, now hosted by the Linux Foundation.

A2A is the telephone network of the Agent world — letting different agents discover, communicate, and collaborate.

Why Do Agents Need Inter-Communication?

Because real-world tasks require multiple specialists. Your "customer service agent" is great at conversation but doesn't understand refund processes. Your "finance agent" knows refunds but doesn't understand what the customer is saying. They need to talk to each other to complete "customer requests a refund."

Four Core Concepts

1. Agent Card (Business Card)

Every A2A agent publishes a JSON "business card" telling the world who they are and what they can do:

// https://finance-agent.example.com/.well-known/agent-card.json
{
  "name": "FinanceAgent",
  "description": "Handles refunds, reconciliation, budget approvals",
  "url": "https://finance-agent.example.com/a2a",
  "capabilities": {
    "streaming": true,
    "pushNotifications": true
  },
  "skills": [
    {
      "id": "process-refund",
      "name": "Process Refund",
      "description": "Process refund requests by order ID and reason"
    }
  ]
}

Analogy: An Agent Card is like a LinkedIn profile — other agents read your Card to decide whether to collaborate with you.

2. Task

The core unit of work in A2A, with a full lifecycle:

submitted → working → completed
                  ↘ input-required (needs more info)
                  ↘ failed

Key features: supports long-running operations, multi-turn interaction, real-time status sync via SSE.

3. Message

Agents exchange information via Messages, each with a role (user / agent) and content (Parts).

4. Artifact

The output when an agent completes a task — can be text, files, or structured data.

A2A Workflow

Customer Service Agent              Finance Agent
    │                                    │
    │ 1. GET /.well-known/agent-card     │
    │───────────────────────────────────→ │
    │ 2. Returns Agent Card              │
    │ ←─────────────────────────────────│
    │                                    │
    │ 3. POST /a2a (create refund Task)  │
    │───────────────────────────────────→ │
    │ 4. Task status: working            │
    │ ←─────────────────────────────────│
    │                                    │
    │ 5. Poll status                     │
    │───────────────────────────────────→ │
    │ 6. completed + Artifact (receipt)  │
    │ ←─────────────────────────────────│

Ecosystem Status

150+ enterprises supporting (Google, Salesforce, SAP, Atlassian...)
Hosted by Linux Foundation
Latest version v0.3 (July 2025)
AWS Bedrock natively supports A2A

MCP vs A2A: Not Rivals — Partners

This is where many people get confused. MCP and A2A are not competitors. They solve completely different problems.

Dimension	MCP	A2A
Problem solved	How agents use tools	How agents collaborate
Direction	Vertical: Agent ↔ Systems	Horizontal: Agent ↔ Agent
Analogy	USB port (connect peripherals)	Phone network (call each other)
Launched by	Anthropic (2024.11)	Google (2025.04)
Protocol	JSON-RPC 2.0	JSON-RPC 2.0 + SSE
Core abstractions	Tools / Resources / Prompts	Agent Card / Task / Artifact
Transparency	Fully transparent (Host sees internals)	Opaque (only knows skills)
Governance	Anthropic-led	Linux Foundation

How They Fit Together

          ┌────────────────────────────────────┐
          │         A2A Protocol Layer          │
          │   Agent ←── collaborate ──→ Agent   │
          │  (Service)                (Finance) │
          └─────┬────────────────────┬─────────┘
                │                    │
          ┌─────▼─────┐      ┌──────▼────┐
          │ MCP Layer │      │ MCP Layer │
          │Agent↔Tools│      │Agent↔Tools│
          └─────┬─────┘      └──────┬────┘
                │                    │
          ┌─────▼─────┐      ┌──────▼────┐
          │  CRM      │      │  ERP      │
          │  Chat logs│      │  Bank API │
          └───────────┘      └───────────┘

MCP is vertical (agents connect down to tools). A2A is horizontal (agents connect to each other). A complete enterprise Agent system needs both.

Practical Example: Cross-Department Procurement

Let's wire everything together with a real scenario — user tells the Procurement Agent: "We need to purchase 200 new laptops."

# Step 1: Procurement Agent uses MCP to query internal systems
# (MCP: Agent → Tools)

inventory = mcp_call("erp_server", "check_inventory",
    {"item": "laptop", "quantity": 200})
# Result: 45 in stock, need to procure 155

budget = mcp_call("finance_server", "check_budget",
    {"department": "IT"})
# Result: remaining budget $72,500

# Step 2: Coordinate with other agents via A2A
# (A2A: Agent → Agent)

supplier_card = a2a_discover("supplier-agent.vendor.com")
# Reads supplier Agent's card: can quote, order, track logistics

quote_task = a2a_send_task(
    agent="supplier-agent.vendor.com",
    skill="get-quote",
    message="155 ThinkPad X1 Carbon, deliver to Shanghai HQ")
# Artifact: unit price $400, total $62,000, delivery 15 days

legal_task = a2a_send_task(
    agent="legal-agent.internal",
    skill="compliance-review",
    message="Supplier quote $62,000, please review compliance")
# Task: working → input-required → completed (approved)

# Step 3: Execute final operations via MCP
# (MCP: Agent → Tools)

po = mcp_call("erp_server", "create_purchase_order", {
    "vendor": "Lenovo",
    "items": [{"name": "ThinkPad X1", "qty": 155, "price": 400}],
    "total": 62000})

mcp_call("slack_server", "send_message", {
    "channel": "#procurement",
    "text": f"PO {po['id']} created, awaiting CFO approval"})

Notice the alternating use of MCP and A2A: MCP handles "downward" operations on specific systems (ERP, Slack), A2A handles "lateral" coordination with other agents (supplier, legal). The entire workflow is automated, with human intervention only at critical points (CFO approval).

What This Means for Developers

What to learn now

Build MCP Servers: The MCP ecosystem is in its "early App Store" phase — the platform exists, but quality apps are severely lacking. One high-quality MCP Server can get massive attention.
Understand A2A Agent Cards: Designing good Agent Cards is like writing good API docs — your agent's discoverability depends on how well the Card is written.
Master LangGraph + MCP + A2A: This is the standard tech stack for building production-grade Agent systems in 2026.

Where the opportunities are

Direction	Opportunity	Skills needed
MCP Server dev	Domain-specific MCP connectors	Python/TS + domain knowledge
A2A platforms	Agent discovery and orchestration	Distributed systems + Agent frameworks
Protocol bridging	MCP ↔ A2A interoperability	Deep knowledge of both protocols
Agentic Commerce	Agents that shop for users	MCP (payments) + A2A (price comparison)
Vertical domains	Healthcare/legal/finance Agent solutions	Protocols + industry expertise

My Practitioner's Take

As someone who contributes PRs to LangChain, runs agents on OpenClaw daily, and has built multi-agent systems with LangGraph:

MCP is already the de facto standard. Regardless of which Agent framework you use, you'll end up calling tools through MCP. Learning MCP has extremely high ROI.
A2A is early but directionally certain. Most multi-agent systems still use framework-internal communication (like LangGraph's state passing), but cross-framework, cross-organization agent collaboration will inevitably need a standard like A2A.
The real barrier isn't the protocols — it's "agent design." Protocols are just communication pipes. How to split agent responsibilities, how to design Agent Cards that others want to use, how to handle trust and errors between agents — these design questions are the real competitive advantage.
Non-English MCP Servers are blue ocean. Current official and community MCP Servers are almost entirely English-world focused. MCP Servers for Chinese users (WeChat, Alipay, Feishu, Chinese academic search) are nearly blank territory. The same applies to French, Dutch, and other language ecosystems.

Looking Ahead: Three-Layer Architecture of the Agent Economy

┌──────────────────────────────────────────────────┐
│              Application Layer                    │
│   Agentic Commerce · Agent Social (Moltbook)     │
│   Enterprise workflow automation · Personal AI    │
├──────────────────────────────────────────────────┤
│              Protocol Layer                       │
│      MCP (tool calls) + A2A (agent comms)        │
│      + UCP/ACP (commerce protocols)              │
├──────────────────────────────────────────────────┤
│           Infrastructure Layer                    │
│    LLM (Claude/GPT/Gemini) + Vector DBs          │
│    + Payments (Skyfire) + Identity/Auth           │
└──────────────────────────────────────────────────┘

In 2026, we're standing at the starting line of the Agent economy. MCP and A2A are like HTTP and TCP/IP in the early internet — they look like mere technical protocols, but they're actually defining the foundational architecture of digital economy for the next decade.

For developers, now is the best time: the protocols are still early, the ecosystem has massive gaps, and those who build first benefit first.

Author: Barbara Wu — Open source contributor to LangChain & Haystack. PhD from Sorbonne University. Studied how language creates meaning, then discovered AI Agents are doing the same thing.

References: