EmblemAI MCP is ready for Claude Code — install in 60 seconds
EmblemAI's hosted MCP server installs in Claude Code with one command. 200+ crypto tools across 7 blockchains, OAuth 2.0 + PKCE, no client registration. Here's what shipped, and why MCP is the right shape for agent wallets.
Written byEmblemAITeam
mcpclaude-codeoauthagent-walletWhat ships today?
EmblemAI's hosted Model Context Protocol server is installable in Claude Code with one command. The install hop takes under a minute end-to-end: run claude mcp add, approve the vault:read scope in the browser, and your agent has 200+ crypto tools across 7 blockchains — Solana, Ethereum, Base, BSC, Polygon, Hedera, Bitcoin.
- 200+ tools: swaps, conditional orders, DeFi positions, NFT operations, cross-chain bridges, market intelligence
- 7 blockchains covered
- OAuth 2.0 + PKCE, API key, and JWT bearer auth all supported
- Read ops fire immediately. Write ops require explicit approval.
How do I install EmblemAI in Claude Code?
One command:
Claude Code opens the /oauth/authorize page at api.emblemvault.ai, you approve the scope, the token gets stored locally, and the server's tools appear in your agent's tool list. No client secret, no manual registration, no configuration file editing.
What unlocked this install path?
The MCP server itself has been running for weeks. What shipped Friday 2026-04-18 in PR #97 on api.emblemvault.ai was the OAuth unlock that lets MCP clients install it without manual client registration.
- OAuth 2.0 discovery endpoints (
/.well-known/oauth-authorization-server,jwks.json) - PKCE public-client authorization-code flow — no client secret required
- RFC 7591 client metadata document support so clients skip dynamic registration
- Loopback redirect URIs (
http://127.0.0.1:<random-port>) for native clients - Hardened authorize UX via Emblem Auth
Why MCP for a crypto wallet?
A wallet SDK you import is a library: you read docs, map functions to tool calls, handle auth, and break when the SDK updates.
A wallet you add via MCP is a capability. Your agent discovers it, OAuths once, and stays current as the server evolves. For agent-side integrations where the user's capability should outlive any particular code release, that's the right shape.
What does the security model look like?
Write operations require explicit user approval per action. Read operations — balances, portfolio, market data — fire immediately. The OAuth scope starts at vault:read and extends on demand.
This matches what agent wallet users actually want: fast answers, slow spends.
Which MCP clients does this work with?
Standards-compliant OAuth 2.0 + PKCE means the install works in every MCP-compliant client:
- Claude Code
- Claude Desktop
- Cursor
- Windsurf
- Gemini CLI
- GitHub Copilot (MCP client mode)
- Any other MCP-compliant client
What's next on the EmblemAI MCP roadmap?
Next drops, in order:
- Transaction tools with opt-out — branch
feat/mcp-transactions, pending merge - Submission to the Official MCP Registry at registry.modelcontextprotocol.io
- Submission to Anthropic's Claude Directory, Cline Marketplace, Smithery, Glama, PulseMCP, mcp.so
- Deeper OAuth implementation notes in our Dev.to walkthrough
Try it
Tell us what you build. Most interesting flow we've seen so far is an agent reconciling a portfolio across Solana and Hedera in one turn — not a thing we thought anyone would do, yet here we are.