Synapse Layer — Trust Infrastructure for AI Agents

io.github.SynapseLayer由 GitHub 账号 SynapseLayer 发布v2.4.6更新于 Oct 24, 2026

MCP-native Trust Infrastructure for AI Agents. Persistent encrypted memory with Trust Quotient.

已验证Streamable HTTP可网页运行知识与记忆AI 与机器学习

概览

AI 生成的概览

远程 MCP 记忆服务,让助手跨会话、跨代理保存、搜索并召回持久化的加密记忆。

可以做什么
  • 保存与存储记忆,包括带元数据和信任评分的结构化条目,以及从自由文本中提取候选记忆
  • 按语义、时间、优先级或混合路由以及全文匹配召回和搜索已存记忆,并提供同义工具别名
  • 为会话初始化持久记忆上下文,并分页列出记忆元数据
  • 提交反馈以调整记忆的信任评分,并在代理之间交接上下文状态
  • 检查服务可用性,获取运行时间与 SLO 指标
适用场景
适合需要助手在会话之间记住用户偏好、事实和既往决策的场景,也适合多个 MCP 客户端与代理共享同一受治理记忆库的情况。团队若需要带信任评分、可搜索的生产代理运行记忆,也可考虑。
运行要求
需要访问 forge.synapselayer.org 的远程端点(streamable HTTP),无需本地安装包;不支持原生 HTTP 的客户端可通过 npx mcp-remote 桥接。需要从 Forge 控制台获取的 Connect 令牌(sk_connect_ 前缀),通过 x-connect-token 请求头发送。需要网络访问。
安装前请注意
x-connect-token 请求头是访问记忆库的密钥凭据,应避免写入共享配置或日志。记忆内容会上传并保存在第三方服务中,不宜存入敏感的个人或商业数据。多个工具会写入、修改或导出数据,部分召回与导出调用需填写理由并受速率限制。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Synapse Layer — Trust Infrastructure for AI Agents,将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Web executable,通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。

交给智能体安装

把这段提示词粘贴到 Claude Code、Codex、Cursor 等智能体,或 SourceWeft 对话中。智能体会读取该服务器的安装指南,先向你展示发布者、端点和登录方式,经你同意后再把它添加到你的客户端。密钥不会写进任何命令。

阅读 https://sourceweft.com/mcp/io.github.SynapseLayer%2Fsynapse-layer/install.md,并按其中说明把这个 MCP 服务器接入我的客户端。修改任何配置之前,先告诉我它的发布者、端点和登录方式,等我确认。不要把 API 密钥写进命令或我会提交的文件;问我密钥应从哪里获取。

查看智能体所依据的安装指南

其他 MCP 客户端

把它添加到你客户端的 mcpServers 配置中。

{
  "mcpServers": {
    "synapse-layer-trust-infrastructure-for-a": {
      "type": "http",
      "url": "https://forge.synapselayer.org/api/mcp"
    }
  }
}

README

🧠 Synapse Layer

RAG retrieves. Synapse remembers.

Persistent memory infrastructure for AI agents — AES-256-GCM encrypted at rest, semantic search, MCP-native.

Synapse Layer is open-source persistent memory infrastructure for AI agents and assistants. Memories are encrypted at rest with AES-256-GCM, indexed via pgvector HNSW for semantic recall, and exposed through MCP JSON-RPC for native integration with Claude, GPT, Gemini, and any MCP-compatible client. Apache 2.0 licensed.

[PyPI] [Python] [Downloads] [MCP Compatible] [Official MCP Registry] [CI] [License: Apache-2.0] [Smithery]

Website · Docs · PyPI · Forge


⚡ 30-Second Quickstart

Get your Connect Token at forge.synapselayer.org → Connect, then install into any MCP client:

bash
curl -fsSL https://forge.synapselayer.org/install/smithery | bash

On the Forge dashboard the Smithery card already copies the command with your token embedded (… | bash -s -- sk_connect_…) — a single paste in your terminal is enough.

Or use the Python SDK:

python
from synapse_layer import Synapse
s = Synapse(token="sk_connect_YOUR_TOKEN")
s.store("user likes coffee")print(s.recall("what does user like?"))

Get your token at forge.synapselayer.org → Dashboard → Connect


What is Synapse Layer?

The persistent memory layer for AI agents — the missing piece between stateless LLMs and real continuity of context.

Your AI agents forget everything between sessions. Synapse Layer fixes that.

FeatureDescription
🔐 Encrypted at restAES-256-GCM with per-operation random IV and HMAC-SHA-256 integrity
🧩 One-click connectClaude Desktop, Cursor, LangChain, CrewAI, n8n
🌐 Cross-agent memorySave in ChatGPT, recall in Claude
⚡ MCP-nativeAny MCP-compatible agent
🔒 Header-first authTokens never in URLs or logs
🎯 Trust QuotientDeterministic recall — memories ranked by confidence, not recency alone

Why Synapse Layer?

Your AI agents forget everything between sessions. Synapse Layer fixes that — in one line.

Without Synapse LayerWith Synapse Layer
Agent forgets context every sessionPersistent memory across all sessions
Memory locked to one modelCross-agent: save in ChatGPT, recall in Claude
No audit trailTrust Quotient scoring on every memory
Complex integrationpip install synapse-layer + 3 lines of code
Plaintext stored on serversAES-256-GCM encrypted at rest

Use Cases

  • Long-term assistant memory — persist user preferences, facts, and prior decisions across sessions.
  • Cross-agent continuity — save context in one agent and recall it in another.
  • Secure memory for MCP clients — connect Claude Desktop, Cursor, and other MCP-compatible tools to a governed memory layer.
  • Operational memory for teams — maintain structured context, trust scoring, and searchable recall for production agents.

Install

bash
pip install synapse-layer

Quick Start

Python Script

python
from synapse_layer import Synapse
client = Synapse(token="sk_connect_YOUR_TOKEN")
# Storeclient.store("User prefers dark mode and concise answers")
# Recallresults = client.recall("user preferences")for r in results:    print(r["content"], r["trust_quotient"])

With Context Manager

python
from synapse_layer import Synapse
with Synapse(token="sk_connect_YOUR_TOKEN") as client:    client.store("User prefers dark mode and concise answers")    results = client.recall("user preferences")    for r in results:        print(r["content"])

Get your token at forge.synapselayer.org → Dashboard → Connect


13 MCP Tools at a Glance

Synapse Layer currently exposes 13 MCP tools for persistent memory workflows:

  • recall
  • save_to_synapse
  • process_text
  • search
  • health_check
  • initialize_context
  • save_memory
  • store_memory
  • recall_memory
  • list_memories
  • memory_feedback
  • neural_handover
  • slo_report

These tools cover memory capture, semantic recall, structured storage, feedback loops, agent handoff, and operational observability.


Deployment Modes

Python Script Mode

Use the SDK when you want direct Python access to Forge memory from your application.

Best for:

  • prototypes and scripts
  • Python-native workflows
  • fast integration into existing apps

Cloud / Forge API

Use Forge when you need persistent, cross-session, and cross-agent memory with managed access tokens.

Best for:

  • production assistants
  • multi-agent systems
  • MCP-based integrations
  • shared memory across tools and sessions

MCP Integration (one tested path per client)

Claude Code (CLI)

bash
claude mcp add --transport http synapse-layer \  https://forge.synapselayer.org/api/mcp \  --header "x-connect-token: sk_connect_YOUR_TOKEN"

The URL is a positional argument — claude mcp add --transport http <name> <url> — there is no --url flag.

Claude Desktop

Add to claude_desktop_config.json:

json
{  "mcpServers": {    "synapse-layer": {      "command": "npx",      "args": [        "mcp-remote",        "https://forge.synapselayer.org/api/mcp",        "--header",        "x-connect-token: sk_connect_YOUR_TOKEN"      ]    }  }}

Cursor

Add to .cursor/mcp.json in your project root (same JSON block as above).

Any MCP client

Any host that runs npx can use the mcp-remote block above unchanged — it bridges local stdio to the remote endpoint.

Config file locations:

  • Claude Desktop — macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Claude Desktop — Windows: %APPDATA%\Claude\claude_desktop_config.json
  • Claude Desktop — Linux: ~/.config/Claude/claude_desktop_config.json

API — Header-First Auth

bash
# Health checkcurl -H "x-connect-token: sk_connect_YOUR_TOKEN" \  https://forge.synapselayer.org/api/connect/health
# Save memorycurl -X POST \  -H "x-connect-token: sk_connect_YOUR_TOKEN" \  -H "Content-Type: application/json" \  -d '{"content": "User is a Python developer"}' \  https://forge.synapselayer.org/api/v1/capture

Security

FeatureImplementation
EncryptionAES-256-GCM at rest with per-operation random IV
IntegrityHMAC-SHA-256 on content
AuthHeader-first (x-connect-token) — tokens never in URLs or logs
PrivacyContent sanitization + tenant-scoped encrypted storage
Isolation1 user = 1 tenant = 1 private mind

See SECURITY.md for vulnerability reporting.


Related Projects

ProjectDescription
synapse-sdk-pythonPython SDK — LangChain, CrewAI, and A2A protocol adapters
synapse-layer-skillMCP skill configuration for Claude Desktop, Cursor, Windsurf
synapse-layer-langgraphLangGraph checkpoint saver with encrypted state persistence

Governance

  • All public claims follow the Public Claims Matrix.
  • Architecture details that reveal benefits are public; mechanisms that enable them are private.
  • Claim = Reality. If it's not implemented, it's not in the README.

License

Apache-2.0 © Synapse Layer

来源:README.md,提交 96d0706

工具

13

recall

Recall Memory

Retrieve relevant persisted memory using semantic, temporal, priority, or hybrid routing. Governance: requires reason (10–200 chars). Rate limit: 20/min.

输入参数modelimitquery必填scopereason必填subjectagent_id

save_to_synapse

Save to Synapse

Persist memory with encryption at rest, sanitization, and deduplication controls.

输入参数tagstypecontent必填projectsubjectagent_idimportanceforce_intentintent_reason

process_text

Process Text

Extract candidate memories from free-form text with governance filters and sanitization.

输入参数text必填sourceprojectagent_id

health_check

Health Check

Check service availability, engine version, and storage health. Public method.

无需输入参数。

initialize_context

Initialize Context

Initialize a persistent memory context for a conversation or agent session. Returns REDACTED metadata by default (detail="summary"). Use detail="full" with a reason (10–200 chars) to retrieve plaintext content under MEMORY_EXPORT_GOVERNANCE_V1.

输入参数detailreason

save_memory

Save Memory

Save a memory entry to the persistent store. Alias of save_to_synapse.

输入参数tagscontent必填subjectagent_idforce_intentintent_reason

store_memory

Store Memory

Store structured memory with metadata and trust scoring.

输入参数tagscontent必填subjectagent_idforce_intentintent_reason

recall_memory

Recall Memory

Recall persisted memory by query. Alias of recall.

输入参数modelimitquery必填scopereason必填subjectagent_id

list_memories

List Memories

List memory metadata with pagination and governance limits.

输入参数limitscopesubjectagent_id

memory_feedback

Memory Feedback

Submit feedback to adjust trust scoring for stored memories.

输入参数signal必填memoryIdmemory_idsessionIdsession_id

neural_handover

Neural Handover

Transfer contextual state between agents with continuity controls.

输入参数token必填reason必填consuming_agent

slo_report

SLO Report

Return uptime and SLO metrics for the MCP service.

输入参数admin_token必填window_hours

版本历史

1
  1. v2.4.6最新Sep 16, 2026

评分

已满足适用检查项的 100%

该评分衡量此收录条目的完整程度,依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价,也不表示服务器安全或当前可用:可连通性是带日期的检查结果,评分也不包含安装量或用户评价。

  • 近 30 天内可连通必需25 分
  • 已声明认证方式必需10 分
  • 已列出工具必需15 分
  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分