# Slant — Agent Setup

> Machine-readable setup for connecting AI agents to Slant (slant.page).
> Human version: https://slant.page/agents
> Product overview for LLMs: https://slant.page/llms.txt
> Full bridge API reference: https://slant.page/docs.txt

## Claude Code

```bash
# MCP (recommended) — the human mints a token in Slant → Settings → AI agents
claude mcp add --transport http slant https://slant.page/api/mcp --header "Authorization: Bearer <token>"

# Skill (publish local markdown + share-link collaboration helpers)
mkdir -p ~/.claude/skills/slant && curl -fsSL https://slant.page/SKILL.md -o ~/.claude/skills/slant/SKILL.md
```

## Codex

```bash
mkdir -p ~/.codex/skills/slant && curl -fsSL https://slant.page/SKILL.md -o ~/.codex/skills/slant/SKILL.md
```

## CLI

```bash
npm install -g slant-cli
slant login                   # opens a browser to approve; or --token / SLANT_TOKEN
slant publish notes.md --title "Notes"
slant read <docId|shareUrl> --json
```

Zero-dependency wrapper over the MCP endpoint. Every command takes `--json`.

## Any MCP client (Cursor, VS Code, ChatGPT, custom)

- Transport: streamable HTTP
- URL: `https://slant.page/api/mcp`
- Auth header: `Authorization: Bearer <token>` (minted in Slant → Settings → AI agents)

A token = scope (own docs / folder / one doc / workspace) × permission (view < comment < suggest < edit) plus optional author powers (create / publish / mint links).

## Share-link bridge (single document, zero install)

Given a share URL `https://slant.page/d/<docId>?token=<token>`, use the REST bridge at `/api/bridge/<docId>/*` with the token as `Authorization: Bearer <token>`. Endpoints: state, edit, delete, comments, suggestions, style, knowledge, presence. Reference: https://slant.page/docs.txt

## Updates

- Skill version: https://slant.page/api/skill-version (the skill's `slant-update` helper checks this)
- The served skill at https://slant.page/SKILL.md is always current.