Skip to main content
This page demonstrates the AGENTS.md documentation template — agent-centric examples, Mintlify components, and standard diagrams.
The user reads a reference page; the hero example and diagram show how feature docs should open with Agent code and a user flow.

Quick Start

1

Agent-Centric Intro

Every page opens with a minimal Agent(...) example before the hero diagram.
2

Mintlify Structure

Wrap Quick Start in Steps, Best Practices in AccordionGroup, and Related links in CardGroup (cols=2).

How It Works


Opting an Example Out of the Guard

The SDK’s session-docs guard executes every self-contained Python example in praisonaiagents/session/README.md; drop <!-- praisonai: skip=true --> on the line above a fence to opt an example out.
The directive is an HTML comment placed within five lines above the fence. Accepted spellings are case-insensitive: skip=true, skip=1, skip=yes.

When to use it

Use the directive when an example:
  • writes to an absolute path outside the temp dir,
  • reads a live secret or credential,
  • starts a long-running server or opens a socket.
For a deliberate fragment, prefer letting the free-name filter drop it — no directive needed.

When not to use it

  • “It doesn’t run in my env” is not a reason. If the example is real, make it hermetic.
  • Don’t use it to hide a broken example. The guard exists so the docs stay honest.
The executor lives at src/praisonai-agents/tests/unit/session/test_session_docs_examples.py, introduced in PraisonAI PR #4179.

Best Practices

Top of every page: a minimal agent example showing the feature in use — not subsystem imports or config-only snippets.
Hero Mermaid diagrams use #8B0000 for agents, #189AB4 for tools/processes, and white text on coloured fills.
Include all imports, use os.getenv() for secrets, and avoid placeholder strings like "your-api-key".
Each section gets a single sentence explaining what follows — no preamble phrases from AGENTS.md §6.3.

Templates

Customise agent prompts and response formatting

Telemetry

Anonymous usage metrics for agent runs