Agent Memory API

Store audit-ready memory after an AI agent makes a decision. Part of Agent Control Primitives — the missing audit memory layer in CDP Bazaar.

A working prototype API for storing and retrieving AI agent memory records with audit-oriented metadata.

Part of AI Agent Infrastructure Safety Stack

This project is part of a small AI-agent infrastructure safety stack.

It focuses on one layer of the emerging problem: how to control autonomous agents before they call APIs, spend money, write memory, or execute external tools.

AI agents are probabilistic. But payments, permissions, memory writes, and external actions require deterministic control.

Related components:

Agent Security Gateway — Prompt injection and policy evasion detection
Agent Budget Guard — Budget and permission check before payment
Agent Memory API — Audit-ready memory storage
Agent Evolution Engine — Orchestration across the stack
Spec-to-Guardrail API — Extract safety rules from API specs

All APIs are indexed in CDP Bazaar and support x402-style payments.

Disclaimer

This is an independent experimental project. It is not officially affiliated with JPYC, Circle, Arc, Kaia, OpenAI, Anthropic, or any related foundation.

This project references JPYC, x402, Arc, Kaia, USDC, and related protocols as technical contexts. It does not imply official partnership, endorsement, or integration unless explicitly stated.

AEO-ready for AI agents

This API is designed to be understandable and callable by AI agents. It includes structured information for:

what this API does
when an AI agent should call this API
when not to use this API
best used before / after
example request / response
next_recommended

Why this matters

AI is moving from answering questions to managing real workflows. As AI agents begin to call paid APIs, use x402 payments, store long-term memory, and interact with other agents, they need structured guardrails before and after execution. This API is part of an AI Agent Payment Safety Stack designed to run as a pre/post-execution control layer on Arc and other agent economies.

Arc compatibility

Designed to run as a post-execution audit and memory storage layer after ERC-8183 job funding on Arc testnet. Call this API after an AI agent funds or executes a job contract on Arc. Compatible with AI agents using USDC, JPYC, x402, or Arc-style payment flows.

What this API does

AIエージェントのセッション間で、支払いポリシー・監査文脈・判断履歴をAES-256暗号化で保存するAPIです。

When an AI agent should call this API

Arc上のERC-8183 job完了後、またはx402決済完了後に監査ログを保存するために呼び出してください。

When not to use this API

一時的な作業メモリの保存
機密情報を暗号化なしで保存したい場合

Best used before

multi-session agent handoff
audit log finalization

Best used after

ERC-8183 job settlement on Arc
x402 payment completion
agent-budget-guard check

Output

memory_id
stored / recalled
audit_log
deletion_proof
next_recommended

Related APIs

Agent Security Gateway
Agent Budget Guard
Agent Evolution Engine

Japanese Agent Trust Layer

このAPIは「Japanese Agent Trust Layer」の一部です。日本語対応AIエージェントが安全・確実・予算内でAPIを使うためのインフラ層を提供します。

Trust Layerの構成

記憶管理: agent-memory-api
安全判定: agent-security-gateway
予算管理: agent-budget-guard
API選定: agent-curator-api
自律進化: agent-evolution-engine

特徴

x402 / USDC決済対応
日本語対応
決定論的バリデーター（AI不使用）
暗号化・削除証跡付き
Base Mainnet対応

⚡ 実装方法

Paid Endpoints (x402 Payment Required)

# AI記憶の保存 (0.05 USDC)
curl -X POST "https://agent-memory-api-bix5.onrender.com/api/memory/store" \
  -H "X-PAYMENT: your-payment-proof" \
  -H "Content-Type: application/json" \
  -d '{
    "agent_id": "your_agent",
    "memory_type": "learning",
    "content": "学習内容",
    "tags": ["重要", "成功パターン"],
    "importance": 0.9
  }'

# 記憶の呼び出し (0.03 USDC)  
curl -X POST "https://agent-memory-api-bix5.onrender.com/api/memory/recall" \
  -H "X-PAYMENT: your-payment-proof" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "成功パターン検索",
    "agent_id": "your_agent",
    "limit": 10
  }'

# 信頼性検証・幻覚検出 (0.20 USDC)
curl -X POST "https://agent-memory-api-bix5.onrender.com/api/trust/verify" \
  -H "X-PAYMENT: your-payment-proof" \
  -H "Content-Type: application/json" \
  -d '{
    "content": "検証対象コンテンツ",
    "verification_level": "strict"
  }'

# コンテキストパッケージ (0.10 USDC)
curl -X POST "https://agent-memory-api-bix5.onrender.com/api/context/package" \
  -H "X-PAYMENT: your-payment-proof" \
  -H "Content-Type: application/json" \
  -d '{
    "project_id": "project_001",
    "handover_type": "complete"
  }'

Free Endpoints

# システムヘルスチェック
curl "https://agent-memory-api-bix5.onrender.com/health"

# データベース統計
curl "https://agent-memory-api-bix5.onrender.com/api/stats"

# x402プロトコル発見
curl "https://agent-memory-api-bix5.onrender.com/.well-known/x402.json"

記憶管理機能

永続的知識蓄積: 学習内容の長期保存
インテリジェント検索: 関連性によるスマート呼び出し
信頼性スコアリング: 情報の品質評価
幻覚検出: 95%精度でAI幻覚を特定
コンテキスト引き継ぎ: プロジェクト情報の完全移行

Installation

Clone repository:

git clone <repository-url>
cd agent_memory_api

Install dependencies:

pip install -r requirements.txt

Configure environment:

cp .env.example .env
# Edit .env with your configuration

Initialize database:

# Ensure PostgreSQL is running
python -c "from database import agent_db; import asyncio; asyncio.run(agent_db.initialize())"

Run server:

python main.py

Environment Variables

Variable	Description	Default
`DATABASE_URL`	PostgreSQL connection URL	Required
`WALLET_ADDRESS`	x402 payment recipient wallet	Required
`TEST_MODE`	Skip payment verification	false
`MAX_MEMORY_LENGTH`	Max characters per memory	10000
`DEFAULT_MEMORY_TTL`	Default memory TTL (seconds)	86400
`PORT`	Server port	8003

Database Schema

memories

Agent memory storage with tagging and expiration
Full-text search capabilities
Embedding support for semantic search

trust_logs

Trust verification results
Hallucination detection analysis
Content reliability scoring

context_packages

Project context handover packages
Multi-level summarization
Memory inclusion tracking

sessions

Agent session tracking
Activity monitoring
Memory count statistics

Usage Examples

Store Memory

curl -X POST "http://localhost:8003/api/memory/store" \
  -H "Content-Type: application/json" \
  -H "X-PAYMENT: {payment_data}" \
  -d '{
    "agent_id": "agent-123",
    "session_id": "session-456",
    "context": "User prefers detailed technical explanations",
    "tags": ["preference", "technical"],
    "ttl": 86400
  }'

Recall Memories

curl -X POST "http://localhost:8003/api/memory/recall" \
  -H "Content-Type: application/json" \
  -H "X-PAYMENT: {payment_data}" \
  -d '{
    "agent_id": "agent-123",
    "query": "technical preferences",
    "limit": 10
  }'

Verify Trust

curl -X POST "http://localhost:8003/api/trust/verify" \
  -H "Content-Type: application/json" \
  -H "X-PAYMENT: {payment_data}" \
  -d '{
    "content": "Recent studies show that AI will replace all jobs by 2025",
    "source_agent": "agent-123",
    "context": "Discussing AI impact on employment"
  }'

Create Context Package

curl -X POST "http://localhost:8003/api/context/package" \
  -H "Content-Type: application/json" \
  -H "X-PAYMENT: {payment_data}" \
  -d '{
    "project_id": "project-xyz",
    "include_memories": true,
    "summary_level": "detailed"
  }'

Payment Protocol

This API uses the x402 payment protocol for monetization:

Network: Base
Currency: USDC
Contract: 0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913

Payment verification includes:

Amount validation
Recipient verification
Transaction hash validation
Replay attack prevention

Architecture

┌─────────────────┐    ┌─────────────────┐
│   FastAPI       │    │  PostgreSQL     │
│   Main Server   │◄──►│   Database      │
└─────────┬───────┘    └─────────────────┘
          │
          ▼
┌─────────────────┐    ┌─────────────────┐
│ Payment         │    │  Memory         │
│ Verifier        │    │  Engine         │
└─────────────────┘    └─────────┬───────┘
                                 │
          ┌─────────────────┐    │
          │  Trust          │    │
          │  Engine         │◄───┤
          └─────────────────┘    │
                                 │
          ┌─────────────────┐    │
          │  Context        │    │
          │  Engine         │◄───┘
          └─────────────────┘

Development

Testing

# Set TEST_MODE=true in .env to skip payment verification
export TEST_MODE=true
python main.py

Database Management

# Initialize database
python -c "from database import agent_db; import asyncio; asyncio.run(agent_db.initialize())"

# Test connection
python -c "from database import agent_db; import asyncio; print(asyncio.run(agent_db.test_connection()))"

# Clean up expired data
python -c "from database import agent_db; import asyncio; asyncio.run(agent_db.cleanup_expired_data())"

Deployment

Railway Deployment

# Install Railway CLI
npm install -g @railway/cli

# Login and deploy
railway login
railway init
railway up

Environment Configuration

Set DATABASE_URL to your PostgreSQL instance
Set WALLET_ADDRESS to your payment wallet
Set TEST_MODE=false for production

Security

Input validation and sanitization
Payment verification and replay protection
Database connection security
Content trust analysis
Memory access controls

Monitoring

Health check endpoint at /health
Database statistics at /api/stats
Comprehensive logging
Error tracking and reporting

License

MIT License - See LICENSE file for details

Support

For issues and questions, please create an issue in the GitHub repository.

AI Agent Safety Stack

Works best with:

Agent Security Gateway: https://agent-security-gateway.onrender.com（危険な命令を止める）
Agent Budget Guard: https://agent-budget-guard.onrender.com（勝手な課金を止める）
Agent Memory API: https://agent-memory-api-bix5.onrender.com（必要な記憶を残す）

Name		Name	Last commit message	Last commit date
Latest commit History 50 Commits
.well-known		.well-known
models		models
.env.example		.env.example
.gitignore		.gitignore
README.md		README.md
ai-agent-policy.json		ai-agent-policy.json
app.py		app.py
context_engine.py		context_engine.py
database.py		database.py
encryption_engine.py		encryption_engine.py
examples.md		examples.md
experiment_engine.py		experiment_engine.py
llms.txt		llms.txt
main.py		main.py
mcp_server.py		mcp_server.py
memory_engine.py		memory_engine.py
openapi.yaml		openapi.yaml
payment_verifier.py		payment_verifier.py
railway.json		railway.json
recall_engine.py		recall_engine.py
requirements.txt		requirements.txt
skill.md		skill.md
test_api.py		test_api.py
trust_engine.py		trust_engine.py

Folders and files

Latest commit

History

Repository files navigation

Agent Memory API

Part of AI Agent Infrastructure Safety Stack

Disclaimer

AEO-ready for AI agents

Why this matters

Arc compatibility

What this API does

When an AI agent should call this API

When not to use this API

Best used before

Best used after

Output

Related APIs

Japanese Agent Trust Layer

Trust Layerの構成

特徴

⚡ 実装方法

Paid Endpoints (x402 Payment Required)

Free Endpoints

記憶管理機能

Installation

Environment Variables

Database Schema

memories

trust_logs

context_packages

sessions

Usage Examples

Store Memory

Recall Memories

Verify Trust

Create Context Package

Payment Protocol

Architecture

Development

Testing

Database Management

Deployment

Railway Deployment

Environment Configuration

Security

Monitoring

License

Support

AI Agent Safety Stack

About

Topics

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages