# Coding Agent Guide

> How to feed Struere docs to coding agents like Claude Code, Cursor, and Copilot

Struere's documentation is optimized for consumption by coding agents. This guide explains how to configure your coding agent to use Struere docs effectively.

## LLM Endpoints

| Endpoint | Content | Best for |
|----------|---------|----------|
| `/llms.txt` | Index of all pages with quick reference | Agent discovery and navigation |
| `/llms-full.txt` | Complete documentation in one file | Comprehensive context loading |
| `/llms-sdk.txt` | SDK section only | Agent/data type/role definition tasks |
| `/llms-api.txt` | API reference only | HTTP integration tasks |
| `/llms-cli.txt` | CLI reference only | CLI command questions |
| `/llms-platform.txt` | Platform concepts only | Architecture and permission questions |
| `/llms-tools.txt` | Tools section only | Tool configuration and templates |
| `/llms-integrations.txt` | Integrations only | WhatsApp, Calendar, Payments setup |
| `/{path}.md` | Individual page as raw markdown | Targeted deep-dives |
| `/openapi.yaml` | OpenAPI 3.1 spec for Chat API | API client generation |

## Developer Skill

The Struere developer skill teaches AI agents behavioral rules, silent failure gotchas, decision frameworks, and workflow checklists. Install it to get automatic guidance when working in a Struere project.

### Install via skills CLI (works with 40+ agents)

```bash
npx skills add MarcoNaik/struere-skill
```

### Install via Struere CLI

```bash
struere init   # auto-installs skill
struere docs   # auto-installs skill
struere pull   # auto-installs skill
```

### Direct access

| Format | URL |
|--------|-----|
| Raw SKILL.md | `https://docs.struere.dev/skill` |
| Zip (for Claude.ai upload) | `https://docs.struere.dev/skill.zip` |

## Claude Code (CLAUDE.md)

Add Struere context to your project's `CLAUDE.md`:

```markdown
## Struere

This project uses the Struere AI agent platform.

- Docs index: https://docs.struere.dev/llms.txt
- Full docs: https://docs.struere.dev/llms-full.txt
- API reference: https://docs.struere.dev/llms-api.txt
- SDK reference: https://docs.struere.dev/llms-sdk.txt
- OpenAPI spec: https://docs.struere.dev/openapi.yaml

Key patterns:
- Chat API: POST /v1/agents/:slug/chat with Bearer token (sk_dev_ or sk_prod_)
- SDK exports: defineAgent, defineData, defineRole, defineTrigger
- Default model: openai/gpt-5-mini
- Package manager: bun
- CLI: struere dev (sync), struere deploy (production)
- Environments: development, production, eval (fully isolated)
```

## Cursor (.cursorrules)

Add to your `.cursorrules` file:

```
This project uses Struere (https://struere.dev), an AI agent platform with built-in data layer, automation, and integrations.

Documentation:
- Full reference: https://docs.struere.dev/llms-full.txt
- API: https://docs.struere.dev/llms-api.txt
- SDK: https://docs.struere.dev/llms-sdk.txt

When working with Struere:
- Chat API: POST /v1/agents/:slug/chat with Bearer token (sk_dev_ or sk_prod_)
- Use defineAgent, defineData, defineRole, defineTrigger from 'struere'
- Default model is openai/gpt-5-mini
- Use bun as the package manager
- Scope rule operators: eq, neq, in, contains (NOT ne)
- Entity link/unlink params: fromId/toId (NOT fromEntityId/toEntityId)
- PolicyConfig has resource, actions, effect (NO priority field)
```

## Common Agent Workflows

### Integrating via API

1. Agent reads `/llms-api.txt` or `/openapi.yaml`
2. Uses slug-based endpoint: `POST /v1/agents/:slug/chat` with Bearer token
3. Handles thread management via `threadId` / `externalThreadId`
4. Handles errors based on status codes

### Creating a new agent

1. Agent reads `/llms-sdk.txt` for `defineAgent` API
2. Scaffolds with `npx struere add agent my-agent`
3. Configures tools, system prompt, and model
4. Syncs with `npx struere dev`

### Setting up permissions

1. Agent reads `/llms-platform.txt` for permission model
2. Creates role with `npx struere add role my-role`
3. Defines policies, scope rules, and field masks
4. Tests with development API key

### Debugging

1. Agent reads the relevant knowledge base article (e.g., `/knowledge-base/how-to-debug-permission-denied.md`)
2. Checks the error pattern against the error codes reference
3. Follows the troubleshooting steps

## Per-Page Raw Markdown

Every documentation page is available as raw markdown at `/{path}.md`. For example:

```
https://docs.struere.dev/introduction.md
https://docs.struere.dev/sdk/define-agent.md
https://docs.struere.dev/platform/permissions.md
https://docs.struere.dev/knowledge-base/how-to-set-up-rbac.md
```

This is useful when an agent needs detailed information about a specific topic without loading the full documentation.