Capabilities Txt Registry

capabilities.txt

A simple, open convention for a website to declare what it can do — the

capabilities an agent can discover and invoke — at a well-known location.

The web taught machines to read in layers. robots.txt says what a crawler *may access*. sitemap.xml says what *exists*. llms.txt says what's *worth reading*. Each answers one narrow question for an automated reader.

None of them answers the question agents now ask: **what can this host actually *do*?**

Agents have stopped only *reading* the web and started *acting* on it. An agent that lands on your site can summarize your docs — but it has no standard way to discover that you expose a "create support ticket" capability, a "check inventory" capability, or "start a return," and how to call them. Today that happens through bespoke, one-vendor-at-a-time integrations.

capabilities.txt is the missing layer: a public, well-known file where a host declares the capabilities it offers, so any agent can discover what it can do.

It is deliberately small. llms.txt worked because you could adopt it in an afternoon. capabilities.txt follows the same rule.

The two forms

A host publishes one or both:

• `/capabilities.txt` — human- and agent-readable markdown: capabilities grouped by category, each with an id, version, and one-line description.
• `/.well-known/capabilities.json` — the structured form: an array of capability references, each resolvable to a full descriptor.

The markdown form is for discovery and reading. The JSON form is for machines that want structure. Publishing the markdown form alone is a perfectly good start.

Format (/capabilities.txt)

# capabilities.txt

One sentence: what this host is and what kind of capabilities it offers.

Structured form: https://example.com/.well-known/capabilities.json

##

( )

• (v ) —
• (v ) —

Rules, kept minimal:

1. Line 1 is # capabilities.txt.
2. A blockquote (>) summary follows: one sentence on what the host offers, plus optional links (the JSON form, docs, an invocation endpoint).
3. `##` headings group capabilities by category (free-form, your choice).
4. `###` headings name a group, optionally with a stable (group-id).
5. Each capability is a list item: a stable capability-id, an optional (v ), and a short — description.
6. It's just markdown. If a human can read it and an agent can parse it, it's valid.

Structured form (/.well-known/capabilities.json)

{

"version": "1", "capabilities": [ { "id": "support.create_ticket", "version": "1.2.0", "description": "Open a support ticket", "descriptor": "https://example.com/.well-known/capabilities/support.create_ticket.json" } ] }

Each entry is a reference; descriptor (optional) points to a full machine-readable description of inputs, permissions, and how to invoke.

Where it sits among the standards

It is not a replacement for MCP or an API spec. The Model Context Protocol is a stateful connection-and-invocation protocol; OpenAPI describes an HTTP API. capabilities.txt is the layer *before* invocation — a static, public, crawlable advertisement an agent (or a search engine) can read with no live connection, that points to your MCP server, HTTP API, or other endpoint for the actual call. Discovery and invocation are different jobs. capabilities.txt does discovery; it hands off invocation.

Adopt it

Fastest (no spec): copy the prompt at capabilitiestxt.org/implement and hand it to your AI coding agent — it writes your capabilities.txt from your code.

Have an OpenAPI spec? Generate it in your browser (paste the URL) — or keep it current automatically in CI with the GitHub Action:

- uses: capabilityhostprotocol/capabilities-txt/action@main

with: openapi: https://api.yoursite.com/openapi.json output: public/capabilities.txt

Then check it at capabilitiestxt.org/submit for a grade, fixes, and a badge — and it’s discoverable in the directory + map.

By hand:

1. List the capabilities your site exposes (or could).
2. Write them into /capabilities.txt using the format above.
3. Optionally publish /.well-known/capabilities.json.
4. Add yourself to adopters.md with a pull request.

Already have an OpenAPI spec? Generate it — no manual authoring:

python tools/from_openapi.py https://api.yoursite.com/openapi.json > capabilities.txt

python tools/validate.py capabilities.txt # check it

Working references — a real, live capabilities.txt plus illustrative templates across markets (e-commerce, support, banking, healthcare, dev platform) — are in examples/. The tools in tools/ generate and validate files.

Where this goes next

Discovery is the first step. Once an agent knows *what* you can do, the next questions are *may I*, *what happened*, and *can I prove it* — invocation, governance, and evidence. Those are defined by the Capability Host Protocol (CHP), an open protocol for which capabilities.txt is the natural public face. You can adopt capabilities.txt on its own; CHP is where it leads if you need the rest.

Status & license

This is a proposal with a working reference, not a finished standard — and it's better for your feedback. Open an issue or PR.

• Specification & site text (README.md, index.html, SPEC.md): CC BY 4.0 — see LICENSE-DOCS.
• Tooling (tools/): Apache-2.0 — see LICENSE.
• "capabilities.txt" is a free, open convention — use it freely. (No trademark.)

Copyright © 2026 Project Auxo, Inc. and contributors.