AESP MCP Server
Enables AI agents to execute vault payments under explicit human control via policy-gated tools (balance checks, deposits, redemptions, transfers) with human review escalation for risky actions.
README
AESP - Agent Economic Sovereignty Protocol
Defining how AI agents operate economically under human sovereignty.
What Is AESP
AESP is a TypeScript SDK and MCP integration layer for agent payments under explicit human control. It is designed so that agents can execute economic actions while humans retain full economic sovereignty.
Core principles:
- Policy-gated execution -- every spend action is bounded by configurable policy rules (per-tx limits, daily/weekly/monthly budgets, address allowlists, time windows).
- Human override path -- risky actions are escalated to a review queue instead of being auto-approved.
- Verifiable commitments -- execution context can be tied to EIP-712 signed intent, enabling on-chain settlement guarantees.
- Practical integration -- MCP tools expose vault operations to AI agent frameworks; subpath exports let you import only what you need.
Install
npm install @yault/aesp
Requires Node.js >= 18.
Quick Start
Add to Claude Desktop
Add the following to your Claude Desktop config (claude_desktop_config.json):
{
"mcpServers": {
"yault": {
"command": "npx",
"args": ["-y", "@yault/aesp"],
"env": {
"YAULT_API_KEY": "sk-yault-..."
}
}
}
}
Add to Claude Code
claude mcp add yault -- npx -y @yault/aesp
Then set the environment variable YAULT_API_KEY in your shell or .claude/settings.json.
Run standalone
export YAULT_API_KEY="sk-yault-..."
yault-mcp # if installed globally
npx @yault/aesp # via npx
Get your API key
Go to yault.xyz to create an account and obtain your API key (sk-yault-*). The key is tied to your vault — each user manages their own key. Agent developers do not need a key unless they are also vault users; it is the end-user who configures their own key in the MCP client.
Use SDK modules
import { PolicyEngine } from '@yault/aesp/policy';
import { getAllMCPTools } from '@yault/aesp/mcp';
import { NegotiationStateMachine } from '@yault/aesp/negotiation';
const engine = new PolicyEngine(storageAdapter);
await engine.load();
const tools = getAllMCPTools(); // 6 MCP tool definitions
Modules
AESP is organized into subpath exports so you can import only what you need:
| Subpath | Description |
|---|---|
@yault/aesp |
Unified re-export of all modules |
@yault/aesp/types |
Shared type definitions (AgentExecutionRequest, TransferPayload, ChainId, etc.) |
@yault/aesp/policy |
Policy engine with 8-check evaluation, budget tracking, and policy change classification |
@yault/aesp/identity |
Agent identity derivation, certificate creation, and hierarchy management |
@yault/aesp/negotiation |
Offer/counter-offer state machine with session management |
@yault/aesp/commitment |
EIP-712 structured commitment builder for dual-signed agreements |
@yault/aesp/review |
Human-in-the-loop review queue with freeze/unfreeze controls |
@yault/aesp/mcp |
MCP tool definitions, argument validation, and server router |
@yault/aesp/a2a |
Agent-card builder for cross-agent discovery (A2A protocol) |
@yault/aesp/crypto |
Cryptographic helpers: signing, encryption, hashing, ZK proof bridge |
@yault/aesp/privacy |
Context tagging, ephemeral address pools, and consolidation scheduling |
MCP Tools
The stdio server exposes 6 backend-connected tools:
| Tool | Method + Endpoint | Purpose |
|---|---|---|
yault_check_balance |
GET /api/vault/balance/:address |
Read a wallet vault balance |
yault_deposit |
POST /api/vault/deposit |
Deposit underlying into vault |
yault_redeem |
POST /api/vault/redeem |
Redeem vault shares |
yault_transfer |
POST /api/vault/transfer |
Transfer vault allocation (parent to sub-account) |
yault_check_authorization |
GET /api/vault/agent-authorization |
Read operator/allowance status |
yault_get_balances |
GET /api/vault/balances/:address |
Read multi-balance breakdown |
Backend Requirements
The MCP server is a thin API client. It expects a Yault backend providing:
GET /api/vault/balance/:addressGET /api/vault/balances/:addressGET /api/vault/agent-authorizationPOST /api/vault/deposit--{ address, amount }POST /api/vault/redeem--{ address, shares }POST /api/vault/transfer--{ from_address, to_address, amount, currency? }
Authentication: Authorization: Bearer sk-yault-* via YAULT_API_KEY env variable.
Security Model
AESP is built around "bounded autonomy":
- Agent API keys should be policy-bound before spend execution.
- Spending controls should include per-tx and rolling limits (daily/weekly/monthly).
- Destination constraints should be allowlist-driven where applicable.
- Sensitive operations should stay outside broad agent key scopes.
- Human escalation remains the fallback for policy violations.
For vulnerability reporting, see SECURITY.md.
Related Packages
| Package | Description |
|---|---|
@yault/elizaos-plugin-aesp |
ElizaOS plugin wrapping AESP for agent frameworks |
Development
Run tests:
npm test
Build TypeScript:
npm run build:ts
Build with WASM (requires acegf-wallet as a sibling repo, or set ACEGF_ROOT):
npm run build:wasm
npm run build:ts
See CONTRIBUTING.md for the full development guide.
License
This project is licensed under the Apache License 2.0. See LICENSE for the full text.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
E2B
Using MCP to run code via e2b.