Agent Paywire library

com.agentpaywire由 agentpaywire.com 发布官方发布者v1.1.0更新于 Oct 24, 2026

Library (L402, x402, Lightning, stablecoins): 18 read-only tools plus 3 writers.

已验证Streamable HTTP可网页运行金融知识与记忆AI 与机器学习

概览

AI 生成的概览

一个远程只读的智能体支付资源库(L402、x402、闪电网络、稳定币),提供来源卡片、文章、数据集、机器人目录、协议事实与付费提醒。

可以做什么
搜索并读取第三方库来源卡片,浏览八个分类,按 id 获取单张卡片。阅读本站文章,并按日期同步来源、文章、数据集、机器人与支持说明的变更。查询带来源引用的智能体支付协议事实与 SDK 安全文档。浏览数据集卡片,以及会付款或连接钱包的公开智能体与机器人目录。检查网址是否含 L402、x402 或 MPP 支付挑战,读取服务商验证状态,列出提醒主题。提交付费路由评价、读取公开评价流,并报告缺失数据。
适用场景
当你需要关于智能体支付协议、闪电网络与稳定币通道的、有来源引用的参考资料,或在集成前确认某网址是否暴露支付挑战时适用。也适合查找数据集、公开付款机器人与 SDK 安全指引。
运行要求
远程 streamable HTTP 端点;未声明任何软件包、账户、API 密钥或请求头。部分工具会返回付费路由的支付信息(例如 2000 sats 或 2.00 USD 的 30 天 webhook),需在工具之外另行付款。
安装前请注意
部分工具会写入:submit_review 会发布评价,request_missing_data 会发送报告,create_alert 只返回支付信息、不会创建订阅。评价以不可信数据存储。付费路由与 webhook 需要花钱。check_payment_endpoint 每个客户端每 24 小时仅两次免费报告,之后只返回支付信息而不执行检查。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Agent Paywire library,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.agentpaywire%2Flibrary/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": {
    "agent-paywire-library": {
      "type": "http",
      "url": "https://mcp.agentpaywire.com/mcp"
    }
  }
}

工具

21

search_sources

Search library sources

Find third-party library source cards. This is the only library lookup: text, exact filters, and newest-first browsing. Output is { count, sources }. Each source has id, title, publisher, authors, date_published (YYYY-MM-DD or null), url, type, category, status, summary, tags, verified_at, and featured only when it is true. count is the rows returned, not the catalog size. query words must all match as case-insensitive substrings of id, title, summary, publisher, type, authors, or tags. category, tag, type, status, featured, and updated_since are AND filters. Omit query to list the newest sources, optionally in one category. Results are newest date_published first, undated last. limit is 1-50 (default 10) with no offset. No matches returns count 0 and sources []. A non-calendar updated_since, or an unknown category, returns an isError result. A status outside the enum is invalid input. A failed library read returns isError whose text is the Error message. Use get_source instead when you already have a source id. Use list_categories instead when you need category ids, labels, and counts rather than source cards. Use list_changes instead when you need the unpaged cross-catalog changelog rather than source cards. Use get_protocol_facts instead when you need cited facts for a payment protocol. Use list_articles instead for this site's own articles. Free, no auth.

输入参数tagtypelimitquerystatuscategoryfeaturedupdated_since

get_source

Get one library source

Fetch one third-party library source card by id. Output is one object: id, title, publisher, authors, date_published (YYYY-MM-DD or null), url, type, category, status, summary, tags, verified_at, and featured only when it is true. Parameter id is required and must match sources[].id from search_sources, or a list_changes entry of kind library (those ids have no prefix). Example: lightning-labs-l402-for-agents. A missing id is rejected as invalid input. An unknown id returns isError with text No source with id <id>. A failed library read returns isError whose text is the Error message. Use search_sources instead when you have no id or need a filtered list (limit 1-50, no offset). Use list_categories instead when you need the category list. Use get_article instead for this site's own articles. Use list_changes instead for changelog rows. Free, no auth.

输入参数id必填

list_categories

List library categories

Return the closed list of eight library categories and nothing else. Output is { categories: [{ id, label, featured, count, json_url, markdown_url, html_url }] }. id is the category id, for example bitcoin-lightning. label is the display name. featured is true only for Bitcoin and Lightning, which is first. count is how many library sources are in that category. json_url, markdown_url, and html_url are absolute links to that category JSON, Markdown, and HTML. There is no source array and no top-level count. Takes no parameters: call with an empty object. The list is unpaged and always the full set; there is no limit or offset. Use list_categories instead of search_sources when you need category ids, labels, or counts. Use search_sources instead when you need the source cards in a category (pass category, for example bitcoin-lightning; limit 1-50, no offset). Use get_source instead when you already have a source id. A failed read of the library file returns isError. The text is the Error message (Fetch failed <status> <url> when the HTTP status is not OK), or Category list failed. when the failure is not an Error. Free, no auth.

无需输入参数。

list_articles

List published articles

List all of this site's own published articles in one unpaged response. Output is { articles: [{ slug, title, summary, date, status, tags, url, md_url }] }. There is no limit or offset. Takes no parameters: call with an empty object. Use get_article instead when you have a slug and need the Markdown. Use search_sources instead for third-party papers, specs, and reports. Use list_changes instead for recent edits (article rows prefix the id with article:). A failed read returns isError. The text is the Error message (Fetch failed <status> <url> when the HTTP status is not OK), or Article list failed. when the failure is not an Error. Free, no auth.

无需输入参数。

get_article

Read one article

Return one of this site's own articles. Output is { slug, markdown }. markdown is the article file with YAML front matter, and the same text is the tool content. Parameter slug is required and must match articles[].slug from list_articles. Example: machine-native-economy. Do not include the article: prefix that list_changes uses. A missing slug is rejected as invalid input. An unknown slug returns isError with text No article with slug <slug>. A failed read returns isError whose text is the Error message (Fetch failed <status> <url> when the HTTP status is not OK), or Article fetch failed. when the failure is not an Error. Use list_articles instead when you do not have a slug. Use get_source instead for a third-party library source. Use search_sources instead to find papers and docs. Free, no auth.

输入参数slug必填

list_changes

List changes after a date

Sync records whose verified_at is later than cursor (exclusive) across library sources, articles, datasets, bots, and support notes. Output is { cursor, count, entries }. cursor is the YYYY-MM-DD that was applied. count is the number of rows in that window. Each entry has kind (article, library, support, dataset, or bot), id, title, url, and verified_at. Some rows also include md_url, category, date_published, or note. Omit cursor to apply an exclusive cursor of UTC today minus 30 days; the response cursor is that date. Pass a YYYY-MM-DD cursor to sync from a stored date. There is no text filter, no limit, and no offset, so an old cursor can return the whole catalog (over 100 entries, about 43 KB). A cursor that is not YYYY-MM-DD returns isError with text cursor must be YYYY-MM-DD. A failed read returns isError whose text is the Error message. This free list is the raw changelog: it is not grouped by protocol and it is not Markdown, and it does not return source cards. Use search_sources instead when you want to browse or filter the library, including newest-first and updated_since (limit 1-50, no offset). Use list_categories, list_articles, list_datasets, list_bots, or get_support_info instead when you want the current full document rather than changelog rows. Use get_trust_checklist instead for the trust checklist. For one record, strip article: before get_article, dataset: before get_dataset, and bot: before get_bot. Library ids have no prefix; use get_source. Support rows have no per-id getter; use get_support_info or get_trust_checklist. Paid GET https://api.agentpaywire.com/v1/changes returns library source cards only, for an inclusive verified_at window, with protocol, category, and new-or-updated filters, as JSON or Markdown. Each client gets 5 unpaid calls per rolling 24 hours. A response in that allowance is HTTP 200 with X-Trial-Remaining and does not spend a credit. After that, the unpaid call is HTTP 402 via x402 or L402. A valid paid call spends one cr

输入参数cursor

get_protocol_facts

Get checked protocol facts

Return source-cited facts for agent payment protocols: spec URL, HTTP status codes and headers, payment rail, networks, facilitators, client libraries, pay flow, and an example. Every fact carries its source URL and verified_at; null means the cited page does not state it. Omit id and the output is the protocol-facts document whose protocols array has all nine protocols (about 23 KB), unpaged. Pass id and the output is that one protocol object. There is no offset. Parameter id is optional and must be one of l402, bolt12, bip21, nwc, x402, mpp, ap2, acp, ucp. Example: l402. An id outside the enum is rejected as invalid input. An enum id missing from the file returns isError with text No protocol with id <id>. A failed read returns isError whose text is the Error message. Use get_protocol_facts instead of search_sources when you need cited protocol facts rather than papers or docs. Use search_sources instead for those papers and docs. Use get_support_info instead for this site's donation addresses and paid API price. Use get_trust_checklist instead for this site's error codes and rate limits. Free, no auth.

输入参数id

get_sdk_security

Read the SDK security registry

Return /api/sdk-security.json. Output is the SDK security document: schema_version, generated_at, disclaimer, description, count, and entries. count equals entries.length. Each entry has id, package, ecosystem (npm or github), minimum_safe_version (string or null), constraint (string or null), per_payment_cap_usd (number or null), formal_advisory_published (boolean or null), advisory_url (string or null), reported_at, fixed_at, origin_header_rejected_since, fixes, notes, source, related_urls, and verified_at. Null means the linked source does not state that field. The rows are @x402/fetch, @x402/axios, @x402/extensions, and @x402/svm at >=2.23.0, Eclair at >=0.14.5, and BTCPay Server with minimum_safe_version null because Optech #428 does not name a version. One unpaged document. Takes no parameters: call with an empty object. Use get_sdk_security instead of search_sources when you need the minimum version table. Use search_sources for the V12 post, Optech #428, and the Eclair release notes. Use /api/lightning-wallets.json and /api/lightning-releases.json for the wallet catalog and the stable node table; this tool does not return those files. A failed read returns isError. The text is the Error message. Free, no auth.

无需输入参数。

get_support_info

Read support and paid-API details

Return the whole voluntary support document from /support.json. Output fields: voluntary (true), purpose, data_is_free (true), access_conditional_on_payment (false), methods, suggested_amounts_sats, suggested_amounts_optional (true), note, updated (YYYY-MM-DD), warning, usdc_contracts_source, and paid_api while the paid API is live. methods[].type is bitcoin-onchain (P2WPKH native-segwit mainnet address, bip21_uri, qr_svg, qr_text), lightning-bolt12 (offer, uri, bip21_uri, qr_svg, qr_text), evm-stablecoin (asset USDC; network Base, Ethereum, or Arbitrum One; chain_id, address, token_contract, eip681_uri; Base is primary and has the QR fields), or solana-stablecoin (asset USDC, network Solana, address, mint, uri, qr_svg, qr_text). suggested_amounts_sats is an array of optional satoshi tip amounts, not a price. paid_api has live, paid, protocol L402, payment Lightning, url, price_sats 100, credits 100, ttl_days 30, and note. One document, unpaged. Takes no parameters: call with an empty object. Calling it never sends money and never creates an invoice. Use get_support_info instead of get_trust_checklist when you need a donation address, a payment URI, a QR link, suggested sat amounts, or the paid_api price block. Use get_trust_checklist instead when you need error codes, the rate-limit policy (no client limit and no HTTP 429; a wallet limit is HTTP 503 with Retry-After), stable URLs, or met/partial checklist status. Use get_protocol_facts instead when you need how a third-party payment protocol works. A failed read returns isError. The text is the Error message (Fetch failed <status> <url> when the HTTP status is not OK), or Support lookup failed. when the failure is not an Error. Free, no auth.

无需输入参数。

get_trust_checklist

Read the trust checklist

Return the whole /trust.json agent trust checklist for agentpaywire.com. Output includes title, description, free_data, items, count (items.length), tools, errors, wallet_error_codes, rate_limits, credits, urls, error_body, usage, and authorization. Each item has id, title, status (met or partial), statement, and evidence URLs. tools lists the advertised MCP tool names. rate_limits.http_429 is false and rate_limits.client is null; a wallet limit is HTTP 503 with Retry-After. credits repeats price_sats, credits, and ttl_days. One unpaged document. Takes no parameters: call with an empty object. Use get_trust_checklist instead of get_support_info when you need error codes, the rate-limit policy, stable URLs, credit terms, or met/partial status. Use get_support_info instead when you need donation addresses, payment URIs, QR links, or the paid_api price block. Use get_protocol_facts instead for third-party protocol facts. A failed read returns isError. The text is the Error message (Fetch failed <status> <url> when the HTTP status is not OK), or Trust checklist failed. when the failure is not an Error. Free, no auth.

无需输入参数。

list_datasets

List dataset cards

List every training and evaluation dataset card in one unpaged response. Output is the /datasets/index.json document: count, model_note, policy, and entries, plus the index title, description, and links. count is the number of cards. Each entries[] item is a full card: id, name, publisher, url, license (string or null), license_verified, license_url, approx_size, size_bytes (integer or null), format, access (open, login, or terms-acceptance), relevance, mirror_allowed, verified_at, http_status, archive_url, data_period, sample_url, and notes. This site's own library dataset is first. Cards are metadata: nothing is downloaded. There is no limit or offset. Takes no parameters: call with an empty object. Use get_dataset instead when you have an id and want that one card from /datasets/<id>/card.json. Use search_sources with category datasets-benchmarks instead for papers and benchmark sources, which are not these cards. Use list_changes instead for recent dataset edits (those ids use a dataset: prefix). A failed read returns isError. The text is the Error message (Fetch failed <status> <url> when the HTTP status is not OK), or Dataset list failed. when the failure is not an Error. Free, no auth.

无需输入参数。

get_dataset

Get one dataset card

Return one dataset card from /datasets/<id>/card.json. Output is one object: id, name, publisher, url, license (string or null), license_verified, license_url (string or null), approx_size, size_bytes (integer or null), format, access (open, login, or terms-acceptance), relevance, mirror_allowed, verified_at, http_status (integer or null), archive_url (string or null), data_period (string or null), sample_url (string or null), and notes. A card is metadata only and does not download dataset files; follow url for the data. Parameter id is required and must match list_datasets entries[].id. Example: agentpaywire-library. Do not include the dataset: prefix that list_changes uses. A missing id is rejected as invalid input. An unknown id returns isError with text Fetch failed 404 <card url>. A failed read for another status uses that same Fetch failed <status> <url> text. Use list_datasets instead when you need the catalog or do not know the id. Use search_sources with category datasets-benchmarks instead for papers about datasets. Free, no auth.

输入参数id必填

list_bots

List verified agents and bots

Return the whole /bots.json catalog of public agents and bots that pay, get paid, or connect a wallet. Output is { schema_version, generated_at, language, title, description, disclaimer, inclusion, schema, html_url, markdown_url, count, entries }. count equals entries.length. disclaimer is "Listed, not endorsed." The inclusion field states the rule: a public repository or demo, a known maker, a push in the 12 months before verification, and a license taken from a LICENSE file or marked unconfirmed. Each entries[] item is the same object get_bot returns: id, name, maker, description, protocols, url, license (string or null), license_verified, license_url (string or null), verified_at (YYYY-MM-DD), http_status (integer or null), archive_url (string or null), funds_or_keys, and notes. funds_or_keys is true when the software can access a wallet or keys. One unpaged response: no limit, no offset, and no text filter. Takes no parameters: call with an empty object. Use list_bots instead of get_bot when you need every entry, the count, the disclaimer, or the inclusion rule. Use get_bot instead when you already have one id and want that single object; an unknown id is an error on get_bot, not an empty list here. Use search_sources instead for papers and docs about agents. Use get_protocol_facts instead for how a payment protocol works. list_changes rows of kind bot prefix the id with bot:; strip that prefix before get_bot. A failed read returns isError. The text is the Error message (Fetch failed <status> <url> when the HTTP status is not OK), or Bot list failed. when the failure is not an Error. Free, no auth.

无需输入参数。

get_bot

Get one verified bot

Return one entry from the bot catalog by exact id. Output is that entry alone, not wrapped in entries or count: id, name, maker, description, protocols (strings), url, license (string or null), license_verified, license_url (string or null), verified_at (YYYY-MM-DD), http_status (integer or null), archive_url (string or null), funds_or_keys, and notes. funds_or_keys is true when the software can access a wallet or keys. Parameter id is required. It must equal list_bots entries[].id. Catalog ids are lowercase letters, digits, and hyphens. Example: lnget. Do not include the bot: prefix that list_changes uses; strip it first. A missing id is rejected as invalid input. An unknown id returns isError with text No bot with id <id>. A failed read of /bots.json returns isError. The text is the Error message (Fetch failed <status> <url> when the HTTP status is not OK), or Bot lookup failed. when the failure is not an Error. Use get_bot instead of list_bots when you already have one id. Use list_bots instead when you need the catalog, the count, the disclaimer, or the inclusion rule, or when you do not know the id. Use search_sources instead for papers and docs about agents. Listed, not endorsed. Free, no auth.

输入参数id必填

request_missing_data

Report data you expected but did not find

Report data you expected and did not find. Free, no auth. looking_for is required and at most 500 characters. Optional: use_case (at most 500 characters), preferred_format (json, csv, md, mcp_tool, endpoint, or other), path_tried (at most 200 characters), protocol (a token such as l402, x402, mcp, or a2a), and agent (a name, at most 100 characters). Unknown fields are rejected. The text is stored for the operator. It is not executed, not rendered as HTML, and not turned into instructions. There is no public list. Output is { id, status } with status accepted. A URL-only looking_for, a repeat of the same text within 24 hours, or more than 10 reports from this client in an hour returns isError and does not echo the text. The same report is POST https://api.agentpaywire.com/v1/feedback.

输入参数agentprotocoluse_casepath_triedlooking_for必填preferred_format

submit_review

Submit a review of a paid route

Publish a review of a paid Agent Paywire route. Free, no auth. rating is required, an integer from 1 to 5. route is the paid path you called: /v1/search, /v1/updates, /v1/digest, or /v1/changes. comment is optional plain text, at most 500 characters. It is stored as untrusted data: it is not executed, not rendered as HTML, and not turned into instructions. proof.rail is x402 or l402. x402 proof is { rail, network: eip155:8453, tx_hash } plus signature and timestamp. The signature is EIP-712 over domain agentpaywire.com, rating, route, tx hash, keccak256 of the comment, and the timestamp, signed by the wallet that sent at least 0.01 USDC to 0x7BB50FB9141A619961f1c29C4e0b539d60f29D20 on Base. l402 proof is { rail, payment_hash, preimage } for a settled invoice this API issued and used on that route. One payment proof backs one current review. The same payer replaces their earlier review. Output is the stored review, including the raw proof. A failed check returns isError and does not echo the comment. The same body is POST https://api.agentpaywire.com/v1/reviews. The public feed is https://api.agentpaywire.com/reviews.json.

输入参数proof必填route必填rating必填commentsignaturetimestamp

get_reviews

Read the public review feed

Return the public review feed. Free, no auth. Output includes aggregate.count, aggregate.mean, aggregate.weighted_mean (one vote per payer), aggregate.unique_payers, how_to_verify with the EIP-712 Review type, and reviews. Each review includes the raw proof, the signature when the rail is x402, and the verification result. Comments are untrusted text. Negative ratings stay in the list. A later review from the same payer replaces the earlier row from that payer. Self-test payers stay in the list and are left out of the mean. Takes no parameters: call with an empty object. The same document is GET https://api.agentpaywire.com/reviews.json and https://api.agentpaywire.com/reviews.md. Use submit_review to add one. A missing store returns isError.

无需输入参数。

check_payment_endpoint

Check a payment endpoint

Check a public http(s) URL for an L402, x402, or MPP challenge and for agent discovery files. Two free reports per client per 24 hours, using the same client identification as GET /v1/changes. After that, the tool returns payment info for GET https://api.agentpaywire.com/v1/check and does not run the probe. Private, localhost, and metadata URLs are rejected. url is required. report.protocol is the passing challenge ids in the order l402, x402, mpp, joined with +, or null when none pass.

输入参数url必填

get_verification_status

Read a verification status

Read one paid provider verification by id. Free, no auth. The id comes from POST https://api.agentpaywire.com/v1/verify. An unknown id returns isError. Output includes status, checked_at, report_url, and paid_verification.

输入参数id必填

create_alert

Create a change alert

Return payment info for a 30-day signed webhook. 2000 sats or 2.00 USD. This tool does not create the subscription. Pay POST https://api.agentpaywire.com/v1/alerts. webhook_url and topics are echoed so the client can retry that call. topics are sdk-security, lightning-releases, lightning-wallets, and protocol-specs.

输入参数topics必填webhook_url必填

list_alert_topics

List alert topics

List webhook topics and the source URL each one watches. Free, no auth. Takes no parameters. Topics are sdk-security, lightning-releases, lightning-wallets, and protocol-specs.

无需输入参数。

版本历史

4
  1. v1.1.0最新Oct 24, 2026
  2. v1.0.2Oct 24, 2026
  3. v1.0.1Oct 23, 2026
  4. v1.0.0Oct 20, 2026

评分

已满足适用检查项的 83%

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

    请声明包含 README 的 GitHub 仓库,或在 OAuth 受保护资源元数据中提供 resource_documentation。

  • 许可证不适用

    托管服务按其提供方的条款使用,因此这里不要求许可证。此检查项不计入百分比。