A3O artifacts
Agent3Opt follows the A3O pattern: machine-readable files that tell AI assistants who you are, what you sell, and how to verify claims—alongside human-readable llms.txt.
Why A3O
Answer engines and agents increasingly rely on structured, crawlable context. A3O artifacts give you a single, explicit place for product facts, policies, and evidence so models are less likely to hallucinate your positioning or miss critical compliance details.
Artifact overview
- llms.txt — Plain-text overview of the site for LLMs (and humans). Agent3Opt publishes public routes such as /llms.txt and /llms-full.txt.
- knowledge.json — Structured product, pricing, and contact facts. See /.well-known/agent/knowledge.json.
- proof.json — Policies, verification, and evidence links. See /.well-known/agent/proof.json.
Publishing paths
For your own site, place equivalents under public/ (or your static host) so they are served at stable URLs. Use the dashboard A3O tools to validate packs and align content with audits.
Notebook A3O tools
a3o.build_pack and a3o.validate_pack help agents generate and check packs against your notebook data.MCP and your repo
MCP is designed for an IDE-first workflow: Agent3Opt returns JSON and proposed patches; your agent applies changes in the repository you have open. See MCP & AI agents for setup and Site audits for how issues become fix tasks.
Search documentation
Jump to a docs page