
Esheria Regulatory Intelligence
io.github.esherialabs由 GitHub 账号 esherialabs 发布v1.2.2更新于 Nov 11, 2026
Citation-backed regulatory intelligence tools for explicit multi-jurisdiction packs
概览
带引用的监管情报服务:让助手按明确的司法辖区包查询义务、罚则、适用性与变更事件。
- 可以做什么
- 检查服务健康与就绪状态,并按辖区列出、查看和比较监管包
- 获取所选包的义务、适用性、罚则、申报日历、证据与审计元数据
- 核验主张、追溯引用,并查询监管条目之间的关系
- 跟踪版本、差异与变更事件,并导出带 trace ID 的有限结果
- 运营者令牌可增加受范围限制的监控、图谱与客户写入操作,每次均需显式确认
- 适用场景
- 当助手需要针对特定辖区和监管包给出有出处的监管答案时使用,例如梳理义务、核对罚则或跟踪长期变化。适合合规、法务运营与审计研究等必须保留引用、版本和局限说明的场景。它不能替代专业法律意见。
- 运行要求
- 可使用带 OAuth 的托管远程端点,或用 uvx 从 esheria Python 包启动本地 stdio 进程。需要仪表盘创建的数据令牌,通过 ESHERIA_API_KEY(或别名 ESHERIA_API_TOKEN)提供,也可作为 bearer 令牌或 X-API-Key 头;ESHERIA_API_BASE_URL 用于设置 API 基址。需要能访问 Esheria API 的网络。
安装
在 SourceWeft 中
- 打开 控制台中的 Esheria Regulatory Intelligence,将其添加到工作区。
- 连接 所需的 凭据。它们会被私密存储,绝不会在公开页面上展示。
- 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.esherialabs%2Fesheria/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。其他 MCP 客户端
把它添加到你客户端的 mcpServers 配置中。 客户端会在首次连接时完成 OAuth 登录。
{
"mcpServers": {
"esheria-regulatory-intelligence": {
"type": "http",
"url": "https://mcp.esheria.ai/mcp"
}
}
}README
Esheria CLI And MCP
Installable command-line and MCP tools for the Esheria Regulatory Pack API.
Global, citation-backed regulatory intelligence for explicit published packs. Discover current readiness, select the intended jurisdiction and pack, and preserve citations, versions, limitations, and trace IDs across CLI, Python, and MCP workflows. Esheria provides regulatory intelligence, not legal advice.
The package exposes two commands:
Version 1.2.2 is the production/stable public release. The CLI, Python
client, and MCP software distributed in the esheria Python package are
licensed under the Apache License 2.0. Hosted API/MCP access, regulatory data,
service outputs, and Esheria trademarks are not licensed under Apache-2.0;
they remain governed by the Esheria Terms of Service,
Privacy Policy, and any applicable customer
agreement. See LICENSE and NOTICE for the exact boundary.
CLI Quickstart
- Create a data API token in the Esheria dashboard.
- Install the command.
pipxis recommended because it keeps command-line tools isolated from project dependencies:
If you do not use pipx, use normal pip:
- Configure your shell. Put these in your terminal for a one-off test, or in
~/.zshrc,~/.bashrc, or your shell profile to keep them:
PowerShell:
ESHERIA_API_TOKEN is also accepted as an alias when ESHERIA_API_KEY is
unset.
Do not commit API keys. The CLI and MCP server read credentials from
environment variables or command-line flags and redact API key values from
diagnostic output. Prefer the environment variable: --api-key can be exposed
through shell history or the operating-system process list.
- Confirm the API is reachable:
- Discover packs, choose a
domain_pack_id, then pass that pack ID to pack-specific commands:
Output flags can be placed globally or on a leaf command:
Use esheria --help and <group> --help to discover the full command tree.
The CLI includes source-watch operations, graph coverage/rebuild operations,
workspace-scoped customer lifecycle commands, and workspace/token/billing
management commands in addition to the read workflows above.
Workspace, token, and billing commands require a management token. Normal
dashboard-created and OAuth connector tokens carry only regulatory:read.
State-changing regulatory workflows require an explicitly created operator
data token with one or more of monitoring:write, graph:write, or
customer:write; regulatory:read alone is rejected. For example:
The dashboard remains the recommended place for self-serve workspace, token, billing, and subscription administration. Keep operator tokens short-lived and grant only the scopes and pack entitlements they require.
The CLI and MCP server are catalog-first: users list packs and then call tools
with the explicit pack ID they want. ESHERIA_DEFAULT_PACK_ID is an optional
client preference, not a server-side jurisdiction default.
The CLI reports the API's readiness labels, limitations, citations, and trace IDs; preserve them in downstream workflows. Published packs may represent a reviewed subset of the full legal corpus, and evaluator-gated claim verification is not available for every pack. Esheria output is regulatory intelligence, not legal advice or a substitute for qualified counsel.
Hosted MCP
Production MCP uses the hosted Esheria endpoint:
Use this endpoint for normal customer onboarding. It avoids local Python, uvx,
virtual environments, and package discovery on the user's machine.
Claude Directory hosts use OAuth. Other agent hosts send a dashboard-created
Esheria data token as a bearer token or X-API-Key. The hosted MCP server
introspects the credential before initialization and calls the Regulatory Pack
API with it, so billing, pack entitlements, trace IDs, and published-only
behavior remain centralized. Invalid and management-only credentials cannot
enumerate tools.
Codex MCP
Set the token where Codex can read it:
Edit ~/.codex/config.toml and add:
Restart Codex, then call esheria_health, esheria_ready, and
esheria_list_packs.
Local MCP Fallback
For local development, or for agent hosts that do not support remote MCP URLs, you can run the stdio server yourself:
Operator-only HTTP transport command:
Production is already deployed at https://mcp.esheria.ai/mcp; end users
should not run this command.
The hosted OAuth profile exposes a read-only 20-tool catalog for health, readiness, pack discovery, obligations, applicability, claim verification, versions, diffs, change events, filing calendars, evidence, penalties, audit metadata, relationship queries, exports, and citation context. Normal API data tokens expose 29 safe read/read-like tools. Operator data tokens add only mutations authorized by monitoring:write, graph:write, and/or customer:write, up to the complete 37-tool catalog. Every mutation
also requires confirm=true; OAuth Directory sessions remain read-only.
The hosted service uses the official MCP SDK and current Streamable HTTP.
Successful tools mirror bounded JSON in text content and structuredContent,
with trace_id and mcp truncation metadata. Use the API or CLI when a
complete large export is needed.
Use stdio only for hosts that do not support remote MCP URLs:
Claude Code
Use Claude Code's remote MCP setup when your installed version exposes it:
- URL:
https://mcp.esheria.ai/mcp - Authorization:
Bearer <client-api-key>
If your Claude Code version only supports local stdio MCP servers, run this fallback once from a terminal:
Then run:
Start or restart Claude Code and ask it to use the Esheria MCP tools.
Claude Desktop
Use Claude Desktop's remote MCP setup when your installed version exposes it:
- URL:
https://mcp.esheria.ai/mcp - Authorization:
Bearer <client-api-key>
If your Claude Desktop version only supports local stdio MCP servers, use the fallback below.
Open the Claude Desktop MCP config file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
Add or merge this object:
Restart Claude Desktop after saving the file.
来源:README.md,提交 e089c9c
工具
0版本历史
1- v1.2.2最新Sep 16, 2026
评分
已满足适用检查项的 85%
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15 分
请让服务器在无凭据时响应 tools/list,或发布可在沙箱中无凭据启动的软件包。
- 描述必需15 分
- 图标10 分
- 文档15 分
- 许可证10 分