Documentation

Install, connect MCP, index projects, and run workflows — operator docs for orkai, the engineer-first AI layer for coding agents.

orkai is the engineer-first AI layer for coding agents — local daemon + CLI so Claude, Cursor, Codex, OpenCode, and other MCP clients follow your engineering process. Workflows, standards, sessions, plans, and indexed code live on your machine.

This is not a hosted memory API for apps you build — it is the process layer you install so every AI tool picks up the same playbook.

How it fits together

License — trial, single-device ($49.99), or three-device ($99).
Daemon — orkai start keeps the daemon running locally; data lives under ~/.orkai/.
MCP — connect your AI assistant with orkai mcp-config.
Projects — orkai init + orkai index make folders searchable (optional).
Review — orkai review validates code or documents against standards you define (optional).

Start here

Quick start — trial → install → start daemon → MCP (six steps)

CLI reference

Page	Commands
Configuration	`~/.orkai/config.yaml`, `config get`, `config set`
Daemon & MCP	`start`, `stop`, `serve`, `status`, `top`, `open`, `logs`, `mcp-config`
Projects & indexing	`init`, `index`, `unindex`, index from URL
Search & browse	`search`, `get`, `list`
Export & import	`export`, `import`
License	`activate`, `deactivate`, `license status`

Features

MCP — connect your AI — Cursor, Claude Desktop, Claude Code, Codex, OpenCode, ChatGPT
Knowledge layer — annotations, @orkai tags, and graph-enriched code for agents
Spreadsheets & data — index CSV and Excel files, query them with SQL
Review — validate code, documents, or CSVs against your standards
Sessions — your agent's memory across conversations
Workflows — visual step-by-step practices your agent follows
Plans, milestones & tasks — structured work tracking across sessions

Run orkai help or orkai help <command> in the terminal for the full flag list on your installed version.