Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion content/guide.json
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@
{ "situation": "You have lots of structured content people and agents want to query", "standards": ["nlweb"] },
{ "situation": "Your bot visits other sites, or you need to verify bots on yours", "standards": ["web-bot-auth"] },
{ "situation": "You want agents to create accounts on your service without a sign-up form", "standards": ["auth-md"] },
{ "situation": "Customers' own agents should sign in and act on their accounts", "standards": ["poppy"], "note": "Track it; Draft 0.1 is from October 9, 2026, and 0.2 is in progress." },
{ "situation": "Customers' own agents should sign in and act on their accounts", "standards": ["poppy"], "note": "Track it; Draft 0.2 is from October 24, 2026, and 0.3 is in progress." },
{ "situation": "You sell goods and want any AI assistant to browse, cart and check out", "standards": ["ucp"] },
{ "situation": "You want your products shown in ChatGPT", "standards": ["acp"] },
{ "situation": "An agent pays and you need proof a person approved those exact terms", "standards": ["ap2"] },
Expand Down
8 changes: 5 additions & 3 deletions content/standards/ai-catalog.md
Original file line number Diff line number Diff line change
Expand Up @@ -11,7 +11,7 @@ analogy: Your site's table of contents for AI.
looks_like: /.well-known/ai-catalog.json
when_to_use: When you publish that same list for MCP Server Cards and older tools. Serve one file at both addresses.
backers: Linux Foundation working group (MCP and A2A communities)
reality_check: Early, with live catalogs at GitHub, Buildkite and Hugging Face. ARD tools look for `ard.json` first and may skip the old address, so serve the same file at both `ai-catalog.json` and `ard.json`.
reality_check: Early, with live catalogs at GitHub, Buildkite, Hugging Face and Cloudflare's developer docs. [The working group's own list](https://github.com/Agent-Card/awesome-ai-catalogs) has 26, mostly small sites. ARD tools look for `ard.json` first and may skip the old address, so serve the same file at both `ai-catalog.json` and `ard.json`.
links:
- label: ai-catalog.io
url: https://ai-catalog.io/
Expand All @@ -24,12 +24,12 @@ related:
note: ARD calls ai-catalog its predecessor. ai-catalog sees itself as the lightweight publishing format; it was split off, not retired.
- id: nlweb
note: A catalog can list an NLWeb endpoint.
last_verified: "2026-10-19"
last_verified: "2026-10-30"
spec_sources:
- kind: git-path
repo: Agent-Card/ai-catalog
path: specification
pin: "e020b2db26f8f13e904a95e0e457457dd3da9efd"
pin: "7ee97d9655a13e1f6442ab505982c03db17134ea"
---

## How it works
Expand All @@ -38,6 +38,8 @@ Like a table of contents, it doesn't contain the chapters; it tells you where ea

A restaurant that works with a booking provider and a jobs provider can list both AI tools, so any assistant that only knows the restaurant's web address can find the booking tool. Catalogs can also point to other catalogs, so a big platform can list many sub-catalogs.

An entry can also carry optional Trust Manifests: claims about the tool, such as who published it or which audits it passed, each of which can be signed. Since October 6, 2026, the [spec](https://ai-catalog.io/spec/) lets several parties, say the publisher and an auditor, each add their own.

## Who's behind it

It's developed in a [Linux Foundation working repository](https://github.com/Agent-Card/ai-catalog) (started October 2025) with members from the MCP and A2A communities, and published at [ai-catalog.io](https://ai-catalog.io/). Even the group's name is inconsistent: ai-catalog.io calls it the "Agent Card Working Group," while Google thanks the "AI Catalog Working Group."
Expand Down
6 changes: 3 additions & 3 deletions content/standards/ard.md
Original file line number Diff line number Diff line change
Expand Up @@ -11,7 +11,7 @@ analogy: A search engine for agent tools.
looks_like: /.well-known/ard.json
when_to_use: When you offer APIs, MCP servers or agents and want agent search registries to find them.
backers: Google-led authors (Google, Hugging Face, R.V. Guha)
reality_check: Early. [Hugging Face runs a public registry](https://huggingface.co/blog/agentic-resource-discovery-launch), and Hugging Face and GitHub now publish catalogs, though GitHub's is only at the older `ai-catalog.json` address. According to [an independent sweep in September 2026](https://github.com/Agent-Card/ai-catalog/issues/115), only 24 of 102 publishers could be found at `ard.json`; the other 78 used only `ai-catalog.json`.
reality_check: Early. [Hugging Face runs a public registry](https://huggingface.co/blog/agentic-resource-discovery-launch), and Hugging Face and GitHub now publish catalogs, though both only at the older [`ai-catalog.json`](https://huggingface.co/.well-known/ai-catalog.json) address. According to [an independent sweep in September 2026](https://github.com/Agent-Card/ai-catalog/issues/115), only 24 of 102 publishers could be found at `ard.json`; the other 78 used only `ai-catalog.json`.
links:
- label: Specification (v0.91)
url: https://agenticresourcediscovery.org/spec/
Expand All @@ -22,12 +22,12 @@ links:
related:
- id: ai-catalog
note: Its predecessor, now the publishing format. ARD is the discovery format. Serve the same file at both addresses.
last_verified: "2026-10-19"
last_verified: "2026-10-30"
spec_sources:
- kind: git-path
repo: ards-project/ard-spec
path: spec
pin: "40151d4a76af2e266723f78d9142f54a89fb5dcf"
pin: "eca2857d0c9b7b27aa2fb72fe3bb69fcd0d18ea9"
---

## How it works
Expand Down
2 changes: 1 addition & 1 deletion content/standards/mpp.md
Original file line number Diff line number Diff line change
Expand Up @@ -33,7 +33,7 @@ spec_sources:

Paying for an online service usually means signing up, getting a key and setting up billing, which an autonomous agent can't easily do. MPP uses the web's long-unused "402 Payment Required" code: the server replies "this costs 2 cents, here's how to pay," the client pays, and the server returns the result plus a receipt.

It supports one-off charges, subscriptions and streaming pay-as-you-go "sessions," with stablecoins on the Tempo blockchain or ordinary cards through Stripe. Picture a research agent paying a fraction of a cent per article from a paywalled archive and settling the total when it finishes.
It supports one-off charges, subscriptions and streaming pay-as-you-go "sessions." One-off charges can use ordinary cards through Stripe; sessions run on stablecoins, such as on the Tempo blockchain. Picture a research agent paying a fraction of a cent per article from a paywalled archive and settling the total when it finishes.

## Who's behind it

Expand Down
24 changes: 12 additions & 12 deletions content/standards/poppy.md
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@ full_name: Personal Agent Protocol (PAP)
job: identity
order: 2
status: early
status_note: Draft 0.1 (Oct 9, 2026), 0.2 in progress; a few small deployments
status_note: Draft 0.2 (Oct 24, 2026), 0.3 in progress; a few small deployments
tagline: acts as you
phrase: Your agent signs in to a company as you
analogy: A valet key for your personal agent.
Expand All @@ -13,9 +13,9 @@ when_to_use: When customers' own agents should sign in and act on their accounts
backers: Sierra + Meta; 7 launch and 35 design partners
reality_check: A draft first published October 9, 2026. A few small companies serve a `poppy.json`, including [Salespeak](https://salespeak.ai/agent-endpoint/), [Lorikeet](https://www.lorikeetcx.ai/blog/personal-agent-protocol), [Ticket Fairy](https://www.ticketfairy.com/.well-known/poppy.json) and [Alpic](https://alpic.ai/.well-known/poppy.json), most for signed-out chats or plain website browsing only. None of the launch partners does yet.
links:
- label: Specification (Draft 0.1)
url: https://personalagentprotocol.org/docs/0.1/spec
- label: Draft 0.2 (in progress)
- label: Specification (Draft 0.2)
url: https://personalagentprotocol.org/docs/0.2/spec
- label: Draft 0.3 (in progress)
url: https://personalagentprotocol.org/docs/spec
- label: Path to 1.0 (open topics)
url: https://personalagentprotocol.org/docs/path-to-1
Expand All @@ -34,29 +34,29 @@ related:
note: Poppy leaves shopping to UCP and ACP, and may overlap with UCP on sign-in.
- id: ap2
note: Payments are not covered yet. Until they are, payment consent needs another protocol, such as AP2.
last_verified: "2026-10-19"
last_verified: "2026-10-30"
spec_sources:
- kind: page
url: https://personalagentprotocol.org/docs/spec
pin: "3b7a81416cfc5f98d4ce1d6f2cb0e359cbf68a46d5a5317b5a4a856855eb50e3"
pin: "24d8da14fd16c944adeb9969221e6d6a6d9ba089f7299ce8c6a8df8a860001f3"
- kind: page
url: https://personalagentprotocol.org/docs/extensions/operations
pin: "6f94925f7306c2d19cb5548e9655c8d414709301ba8ef37c6e1a436969614b7f"
pin: "ec14d1af5ac4a65e0f691e49fbdf57ede32f32552d86ed88865e85bf8eedb169"
- kind: page
url: https://personalagentprotocol.org/docs/guides
pin: "f2b0a17c0f420732a721f2843717015cb3f8277fc365a6b829af5e36bd67ea49"
pin: "708fc192545d0a65b596c8dbee038338fe559f21f4d3c203023304a1168420d0"
---

## How it works

It lets the agent act for you at a company, with only the permissions you gave it, without opening the boot or the glovebox. Right now an AI agent acting for you has two bad options: pretend to be you in a browser, or not get in at all.

Poppy gives the company a file at `/.well-known/poppy.json` that says how agents can start a session, how you sign in, and which web pages, APIs and company agent are available. Sign-in uses OAuth, the familiar "Sign in with..." approval screen: you approve on the company's own page, on another device with a code, or (if the company allows it) the agent signs in for you. An agent can also start signed out, as a guest.
Poppy gives the company a file at `/.well-known/poppy.json` that says how agents can start a session, how you sign in, and which web pages, APIs and company agent are available. Sign-in uses OAuth, the familiar "Sign in with..." approval screen: you approve on the company's own page, on another device with a code, or (if the company allows it) the agent signs in for you. [Draft 0.3](https://personalagentprotocol.org/docs/guides) calls that last way temporary, with an end date to be set at version 1.0. An agent can also start signed out, as a guest.

Permissions are simple scopes like "read" and "write," plus custom ones such as "manage saved addresses." You also get a different anonymous ID at each company, so companies can't match you with each other. In the launch demo, Meta's Muse agent found Rocket Mortgage's agent and, with the user's permission, completed a mortgage pre-approval.
Permissions are scopes. Version 0.2 has two shared ones, "read" and "write," and companies can add their own, such as "manage saved addresses." [Draft 0.3](https://personalagentprotocol.org/docs/changelog) drops the shared ones: each company names its own and says which make changes that can't be undone. For an action such as exchanging an item, the company can also [ask you to approve the exact terms](https://personalagentprotocol.org/docs/extensions/operations) before it acts. You also get a different anonymous ID at each company, so companies can't match you with each other. In the launch demo, Meta's Muse agent found Rocket Mortgage's agent and, with the user's permission, completed a mortgage pre-approval.

## Who's behind it

[Sierra](https://sierra.ai/blog) (led by Bret Taylor and Clay Bavor) and Meta announced it on October 6, 2026. [Sierra published Draft 0.1](https://sierra.ai/blog/poppy) on October 9 with seven launch partners (Genesys, Instinct, NiCE, Rocket, Shopify, Stripe and Walmart) and 35 more design partners, including OpenAI, Visa, Mastercard and PayPal. [Draft 0.2](https://personalagentprotocol.org/docs/changelog) is in progress, with no publication date yet.
[Sierra](https://sierra.ai/blog) (led by Bret Taylor and Clay Bavor) and Meta announced it on October 6, 2026. [Sierra published Draft 0.1](https://sierra.ai/blog/poppy) on October 9 with seven launch partners (Genesys, Instinct, NiCE, Rocket, Shopify, Stripe and Walmart) and 35 more design partners, including OpenAI, Visa, Mastercard and PayPal. [Version 0.2](https://personalagentprotocol.org/docs/0.2/spec) followed on October 24, and [draft 0.3](https://personalagentprotocol.org/docs/changelog) is in progress.

The spec warns that any part can change, even in ways that break compatibility, and [lists payments as not yet covered](https://personalagentprotocol.org/docs/path-to-1). A reference implementation was promised within about a month of October 9 and isn't out yet. An implementation kit for companies was added on October 17: a skill that has a coding agent design and build a company's side, and a validator that acts as a test agent against a deployment. It reuses OAuth, OpenAPI and MCP rather than replacing them.
The spec warns that any part can change, even in ways that break compatibility, and [lists payments as not yet covered](https://personalagentprotocol.org/docs/path-to-1). A reference implementation was promised within about a month of October 9 and isn't out yet. An implementation kit for companies was added on October 17: a skill that has a coding agent design and build a company's side, and a validator that acts as a test agent against a deployment. [The kit](https://personalagentprotocol.org/kit/company) now follows draft 0.3, not the published 0.2. Poppy reuses OAuth, OpenAPI and MCP rather than replacing them.
12 changes: 6 additions & 6 deletions content/standards/web-bot-auth.md
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,7 @@ name: Web Bot Auth
job: identity
order: 1
status: production
status_note: WG draft -00 (Sept 1, 2026); live signers
status_note: WG draft -01 (Oct 29, 2026); live signers
tagline: which bot is it
phrase: Cryptographic ID for bots
analogy: A tamper-proof ID badge for bots, instead of a name tag anyone could write.
Expand All @@ -23,23 +23,23 @@ related:
note: auth.md signs up a new user for the agent; Web Bot Auth only proves which bot is asking.
- id: x402
note: Cloudflare plans to use them together for pay-per-crawl.
last_verified: "2026-10-19"
last_verified: "2026-10-30"
spec_sources:
- kind: ietf-draft
name: draft-ietf-webbotauth-httpsig-protocol
pin: "00"
pin: "01"
---

## How it works

Today websites identify bots by a label (the "user agent") that anyone can fake, or by IP address, which breaks easily. With Web Bot Auth, the company running the bot publishes a public key at a well-known address and signs every request with the matching private key, using an existing IETF standard for signing web requests (RFC 9421).
Today websites identify bots by a label (the "user agent") that anyone can fake, or by IP address, which breaks easily. With Web Bot Auth, the company running the bot publishes its public keys at a web address, names that address in each request, and signs the request with the matching private key, using an existing IETF standard for signing web requests (RFC 9421).

The website, or its CDN such as Cloudflare, checks the signature and knows for certain "this really is OpenAI's agent" before deciding to let it in, block it or charge it. A news site can let a verified shopping agent through while stopping an anonymous scraper wearing the same label.
The website, or its CDN such as Cloudflare, checks the signature and knows for certain "this really is OpenAI's agent" before deciding to let it in, block it or charge it. A site can also ask an unsigned bot to sign. A news site can let a verified shopping agent through while stopping an anonymous scraper wearing the same label.

It answers *which agent* is visiting, not *whom it represents*; that's Poppy's and OAuth's job.

## Who's behind it

Cloudflare proposed it in May 2025. On September 1, 2026 the IETF's webbotauth working group [adopted it as an official working-group draft](https://datatracker.ietf.org/doc/draft-ietf-webbotauth-httpsig-protocol/), written by authors from Cloudflare and Google; it's the group's only adopted document. It's still a draft, not an RFC: adoption isn't approval.

[Visa's Trusted Agent Protocol and Mastercard's Agent Pay build on it](https://blog.cloudflare.com/secure-agentic-commerce/), adding a tag that says whether the agent is browsing or buying, and observers report OpenAI, Cloudflare and DuckDuckGo signing requests today. The working-group draft has already changed how a request points to its key directory, but [Cloudflare's own checker](https://developers.cloudflare.com/bots/reference/bot-verification/web-bot-auth/) still expects the older format (as of October 8, 2026), so early implementations may need updates.
[Visa's Trusted Agent Protocol and Mastercard's Agent Pay build on it](https://blog.cloudflare.com/secure-agentic-commerce/), adding a tag that says whether the agent is browsing or buying. [OpenAI's ChatGPT agent](https://help.openai.com/en/articles/11845367), [Cloudflare's Browser Run](https://developers.cloudflare.com/browser-run/reference/automatic-request-headers/) and, as an experiment, [some Google-Agent traffic](https://developers.google.com/crawling/docs/crawlers-fetchers/web-bot-auth) sign requests today. [Revision -01](https://datatracker.ietf.org/doc/html/draft-ietf-webbotauth-httpsig-protocol-01) (October 29, 2026) changed how a request points to its keys: it now gives the exact address of its key set. [Cloudflare's own checker](https://developers.cloudflare.com/bots/reference/bot-verification/web-bot-auth/) still expects the older format (as of October 8, 2026), so early implementations will need updates.
Loading
Loading