# agents.txt

> A proposed root-level file that declares a site's identity, terms of use, service catalog and agentic endpoints to visiting AI agents.

_The Agentic Web Lexicon · /glossary/agents-txt · [JSON](/api/glossary/agents-txt) · [all The Agentic Web Lexicon](/glossary)_

- **term:** agents.txt
- **category:** protocols
- **short_def:** A proposed root-level file that declares a site's identity, terms of use, service catalog and agentic endpoints to visiting AI agents.
- **long_def:** Where robots.txt answers 'may you look at this?', agents.txt aims to answer 'what can you do here, and on what terms?' — a B2A (business-to-agent) capability manifest. The namespace is contested: several independent projects have used the agents.txt name for different purposes, and one prominent proposal was renamed agent-manifest.txt in March 2026 to disambiguate. It is an emerging, not-yet-standardized convention.
- **see_also:** agents-md, robots-txt, llms-txt
- **etymology_origin:** — (verify-against-primary-at-build)
- **related_to:** agents-md, robots-txt, llms-txt, agent-identity
- **contrast_with:** Unlike robots.txt, an exclusion contract that says what crawlers may NOT fetch, agents.txt is meant to advertise what agents CAN do — a capability and terms manifest rather than an access-denial list.
- **example:** By March 2026 the agents.txt namespace was crowded enough that one prominent proposal was renamed agent-manifest.txt to disambiguate it from competing uses.
- **source:** https://github.com/asturwebs/agents-txt
- **status:** proposed
- **why_it_matters:** agents.txt represents the emerging move from blocking agents (robots.txt) to inviting and instructing them — but its contested namespace means a site should track which proposal wins before committing.
- **sameAs:** —
- **bridge_entity:** protocols/discovery/agents-txt
- **last_verified:** 2026-06-15
- **md_twin:** /glossary/agents-txt.md