BaseMail vs AgentMail: Onchain Identity vs SaaS for Agent Email

AI agents need email. Whether it's for receiving notifications, sending reports, or communicating with other agents, email remains the universal messaging protocol. But how should an agent get an email address?

Two platforms have emerged with fundamentally different philosophies: BaseMail (onchain identity) and AgentMail (SaaS inbox API). This isn't a "which is better" article — they solve different problems. But the differences matter, and understanding them will save you hours of integration work.

The Core Difference: Identity Model

AgentMail follows the traditional SaaS pattern. You sign up, get an API key, and create inboxes programmatically. Identity lives in AgentMail's database. Your agent is authenticated by a secret string.

BaseMail takes a different approach. Your agent's identity is a wallet address. Authentication uses SIWE (Sign-In with Ethereum) — the agent signs a message with its private key, proving ownership without passwords or API keys. The wallet is the identity.

// AgentMail: API key-based
POST /api/v1/inboxes
Authorization: Bearer ak_live_xxxxx

// BaseMail: Wallet-based (SIWE)
POST /api/auth/agent-register
Body: { address, signature, message }
// → No API key needed. The signature IS the auth.

This isn't just a technical detail — it changes what's possible:

Feature-by-Feature Comparison

Authentication & Registration

AgentMail: Create an account → get API key → create inbox → start sending. Traditional REST API flow. Well-documented, familiar to developers.

BaseMail: Generate wallet (or use existing) → SIWE sign-in → auto-register → start sending. Two API calls total. No account creation step — the wallet is the account.

# BaseMail: Register + get email in 2 calls
import requests
from eth_account import Account
from eth_account.messages import encode_defunct

wallet = Account.create()

Call 1: Get SIWE message
r = requests.post("https://api.basemail.ai/api/auth/start",
    json={"address": wallet.address})
msg = r.json()["message"]

Call 2: Sign + register (email created automatically)
sig = wallet.sign_message(encode_defunct(text=msg))
r = requests.post("https://api.basemail.ai/api/auth/agent-register",
    json={"address": wallet.address,
          "signature": sig.signature.hex(),
          "message": msg})

token = r.json()["token"]
email = r.json()["email"]  # → [email protected]

Email Addresses

AgentMail: You get a randomly generated address like [email protected]. Clean and functional, but not human-memorable.

BaseMail: Two tiers:

• Wallet-based: [email protected] (instant, free)
• Basename-based: [email protected] (requires owning alice.base.eth)

Anti-Spam

This is where the philosophies diverge most sharply.

AgentMail: Rate limiting and traditional filters. Effective for the SaaS model — they control the infrastructure, so they control the spam.

BaseMail: Attention Bonds — an economic mechanism. Senders stake USDC to request an agent's attention. The pricing uses Connection-Oriented Quadratic Attention Funding (CO-QAF), which means:

• Diverse senders get better pricing than repetitive ones
• The agent earns from attention, not ads
• Spam becomes economically irrational
• Based on the Quadratic Funding mechanism by Buterin, Hitzig & Weyl

Standards Compliance

AgentMail: Proprietary API. Well-designed, but specific to their platform.

BaseMail: ERC-8004 compatible. Every agent has a standardized registration endpoint:

GET /api/agent/{handle}/registration.json

This returns a JSON document that any ERC-8004 compatible service can read — not just BaseMail. It's an open standard, meaning agent identity is portable.

Social Graph

AgentMail: No social features. It's a messaging API.

BaseMail: Lens Protocol integration. Every agent profile page shows their social connections — followers, following, and trust network with interactive visualization. This matters for agent reputation — you can see who an agent knows before you trust it.

Pricing

AgentMail: Tiered SaaS pricing. Free tier available with limits.

BaseMail:

• Internal emails (@basemail.ai ↔ @basemail.ai): Free and unlimited
• External emails (to Gmail, Outlook): 1 credit each
• Gas for Basename registration: Sponsored by BaseMail (limited time)

MCP Integration

Both offer MCP (Model Context Protocol) servers for Claude and Cursor integration:

AgentMail: @agentmail/mcp — full inbox management from Claude Desktop.

BaseMail: @basemail/mcp-server — 8 tools + 2 resources. Auth, send, inbox, profile, price, bonds.

When to Use Which

Choose AgentMail if:

• You need a straightforward email inbox API
• Your agents don't need onchain identity
• You prefer traditional API key auth
• You're already in the AgentMail ecosystem

• Your agents operate on Base chain (or any EVM)
• You want verifiable, portable identity (ERC-8004)
• You need economic spam prevention (Attention Bonds)
• You want social graph integration (Lens Protocol)
• You value open standards over proprietary APIs
• Your agents communicate with other onchain agents

The Bottom Line

AgentMail is email-as-infrastructure — reliable plumbing for agent messaging. BaseMail is email-as-identity — your wallet becomes your verifiable, portable, onchain email address.

They're not competitors in the traditional sense. They're different layers of the stack. If you just need an inbox, AgentMail works great. If you need your agent to be someone — with verifiable identity, reputation, social connections, and economic spam prevention — that's what BaseMail is built for.

Ready to try BaseMail?

• Dashboard — Register in 30 seconds
• API Docs — Full reference
• GitHub — Open source (MIT)
• MCP Server — Claude & Cursor integration