Deploy to Agents

com.deploytoagents由 deploytoagents.com 发布官方发布者v0.3.0更新于 Oct 25, 2026

Audit agent-distribution surfaces and create an evidence-based distribution plan.

已验证Streamable HTTP可网页运行网页搜索与抓取开发者工具

概览

AI 生成的概览

远程 MCP 服务器,审计公开应用的面向智能体的表面,并把结果转化为按优先级排序的分发计划。

可以做什么
对公开的 HTTP 或 HTTPS 应用排队执行一次有范围限制的审计,并返回一个不公开列出的回执链接,然后反复查询结果直到处理完成,查看状态、评分、发现项、隐私状态和证据链接。把已完成的技术审计转化为按优先级排序的修复与分发计划,涵盖官方注册表、公开代码仓库、客户证据和独立发现。获取服务方自身的公开自审计证据,包括技术就绪度与独立发现之间明确划定的边界。
适用场景
当你希望助手评估某个公开网站或应用面向智能体分发的就绪程度,并依据审计发现起草具体、有证据支撑的修复与上架计划时适用。
运行要求
远程 Streamable HTTP 端点,无需本地运行时或安装包。README 说明服务器会验证登录组织是否拥有目标主机名,因此审计隐含需要账号并登录。
安装前请注意
审计针对公开的 HTTP 或 HTTPS 应用,并生成不公开列出的回执链接,应把这些链接视为敏感信息。README 说明该服务不保证任何模型会推荐,独立发现也尚未得到证实,不要把评分当作背书。README 还说明它避免虚假证言、自动发帖和操纵排名。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Deploy to 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/com.deploytoagents%2Fserver/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": {
    "deploy-to-agents": {
      "type": "http",
      "url": "https://deploytoagents.com/mcp"
    }
  }
}

README

Deploy to Agents SDK

Public JavaScript client and registry metadata for the Deploy to Agents remote MCP server.

The remote server is published in the official MCP Registry as com.deploytoagents/server and is served from https://deploytoagents.com/mcp. The JavaScript client and authenticated agent-first CLI are published as [email protected] on npm, with an equivalent deploytoagents==0.1.1 client on PyPI.

Agent-first CLI

bash
npx deploytoagents loginnpx deploytoagents whoaminpx deploytoagents portfolio --jsonnpx deploytoagents discovery --jsonnpx deploytoagents discovery-record --input observation.json --jsonnpx deploytoagents audit https://example.com --json

The package installs both deploytoagents and the shorter d2a command. Google login uses Authorization Code with PKCE and a temporary loopback callback. On Windows the refresh credential is encrypted for the current OS user with DPAPI; CI can provide a short-lived identity token through DEPLOYTOAGENTS_TOKEN. Discovery Lab can be read or supplied with a JSON observation file (or stdin via --input -). All command results and errors have stable JSON forms for agent use.

json
{  "hostname": "example.com",  "surface": "claude",  "model": "model label shown by the surface",  "prompt": "Exact generic, unbranded prompt",  "outcome": "not-mentioned",  "freshSession": true,  "responseExcerpt": "Optional relevant excerpt",  "citations": ["https://example.org/source"]}

Valid outcomes are recommended, mentioned, not-mentioned, and error. The server verifies that the signed-in organization owns the target hostname.

Deploy to Agents currently audits public agent-facing surfaces, returns unlisted evidence receipts, and creates prioritized technical and external-authority distribution plans. It does not yet claim to publish every customer artifact or guarantee recommendation by any model.

Connect directly through MCP

Use this Streamable HTTP endpoint in any compatible MCP client:

text
https://deploytoagents.com/mcp

Available tools:

  • audit_app
  • get_audit_result
  • create_distribution_plan
  • get_customer_zero_evidence

JavaScript client

bash
npm install deploytoagents
js
import { DeployToAgentsClient } from "deploytoagents";
const client = new DeployToAgentsClient();
try {  const queued = await client.auditApp("https://example.com");  console.log(queued.receipt_url);
  const result = await client.getAuditResult(queued.audit_id);  if (result.status === "completed") {    console.log(await client.createDistributionPlan(queued.audit_id));  }} finally {  await client.close();}

Python client

bash
pip install deploytoagents
python
from deploytoagents import DeployToAgentsClient
async with DeployToAgentsClient() as client:    queued = await client.audit_app("https://example.com")    print(queued["receipt_url"])

Customer Zero

Deploy to Agents uses its own system as its first customer. The current 100/100 receipt verifies technical surfaces only. Independent, unbranded discovery remains explicitly not-yet-proven.

Development

bash
npm installnpm testnpm run smoke

Evidence policy

This project distinguishes owned evidence, externally verified artifacts, and independent recommendations. It does not create fake testimonials, automated community posts, coordinated votes, or links intended primarily to manipulate rankings.

License

MIT

来源:README.md,提交 52d35c2

工具

4

audit_app

Audit an app for agent distribution

Queue a bounded audit of a public HTTP or HTTPS app and return an unlisted receipt URL. Reuse get_audit_result until processing completes.

输入参数url必填

get_audit_result

Get an audit result

Retrieve the current status, score, findings, privacy state, and evidence URL for an audit UUID.

输入参数audit_id必填

create_distribution_plan

Create an evidence-based distribution plan

Turn a completed technical audit into prioritized remediation plus official-registry, public-repository, customer-evidence, and independent-discovery actions.

输入参数audit_id必填

get_customer_zero_evidence

Get Deploy to Agents Customer Zero evidence

Return the public self-audit evidence and the explicit boundary between technical readiness and independent discovery.

无需输入参数。

版本历史

1
  1. v0.3.0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

其他 Deploytoagents 集成