## What is agents.txt?

agents.txt defines access policies for AI agents — what they can do, how fast, and any restrictions. Think of it as [robots.txt](/kb/robots-txt) evolved for autonomous agents that don't just crawl but take actions.

## Example

Serve at `/agents.txt`:

```
[agent:*]
Allow-Actions: read, search
Rate-Limit: 60/minute

[agent:claude]
Allow-Actions: read, search, post, vote
Rate-Limit: 120/minute

[agent:gptbot]
Allow-Actions: read
Rate-Limit: 30/minute
```

## Key concepts

- **Agent sections**: Define policies per agent or for all agents (`agent:*`)
- **Allow-Actions**: What actions agents can take (read, write, post, etc.)
- **Rate-Limit**: How many requests per time period
- **Agent-specific rules**: Different agents can have different permissions

## Spec maturity

**Emerging standard — not yet formal.** Multiple proposals exist:
- An [IETF Internet-Draft](https://www.ietf.org/archive/id/draft-srijal-agents-policy-00.html) defines a formal policy file format
- Community proposals on GitHub define alternative formats

The concept is gaining traction but the format is not finalized. The example above follows the most common community convention.

## Learn more

- [IETF agents-policy draft](https://www.ietf.org/archive/id/draft-srijal-agents-policy-00.html) — Formal specification proposal

## Related

- [llms.txt](/kb/llms-txt)
- [SKILL.md](/kb/skills)
- [OpenAPI](/kb/openapi)