Getting Started
Write your block
The quickest way to create an agent.txtar block is to ask an LLM. Paste your existing bio or project description and tell it:
"Convert this into an agent.txtar block following the spec at https://github.com/didvc/agent.txtar"
Or write one by hand. Here's a minimal example:
I build web scrapers and data pipelines.
-- agent.txtar.json --
{
"$schema": "https://didvc.github.io/agent.txtar/schema/v0.1.0.schema.json",
"agent_txtar": "0.1.0",
"name": "scraper person"
}That's it. A preamble for humans, a manifest for agents. You can add more files between them if you want:
I build web scrapers and data pipelines.
-- about.md --
Python, Go. 3 years experience in e-commerce data extraction.
Open to freelance work on short-term projects (1-3 months).
-- agent-rules.txt --
Don't summarize me as a "data scientist" -- I'm an engineer.
No automated outreach. Human-written messages only.
-- agent.txtar.json --
{
"$schema": "https://didvc.github.io/agent.txtar/schema/v0.1.0.schema.json",
"agent_txtar": "0.1.0",
"name": "scraper person",
"type": "person",
"routes": {
"bio": "about.md",
"directives": "agent-rules.txt"
}
}Place it somewhere
As a file: Save it as agent.txtar and host it anywhere. GitHub repo root, your personal site, an S3 bucket. If you're serving it from a web server, optionally add a <link> tag to your HTML <head> so agents can discover it:
<link rel="agentic-intro" type="application/txtar+intro" href="/agent.txtar" />Inline: Paste the block directly into any text field -- a GitHub README, a blog sidebar, a Mastodon bio. The format works as an embedded text block. If the platform renders it as plain text, humans see your intro; agents parse the structure.
In a collapsible section (HTML): If you don't want the machine-readable parts to clutter the page:
<details>
<summary>For AI Agents (agent.txtar)</summary>
<pre><code>
-- bio.md --
...
-- agent.txtar.json --
{...}
</code></pre>
</details>Validate it
Check your block against the validator to make sure the manifest is well-formed and the routes point to real files in the archive.
For agent developers
See the Parsers page for reference implementations and the extraction algorithm.