Documentation: set up FlyDocs and run your first workflow

Get started with FlyDocs: install the CLI, connect Linear or Jira, configure skills, and run your first capture-to-close workflow.

Beta

For developers: your AI agent handles the overhead (issue tracking, status updates, formatting, context loading) so you stay in flow and write code. For teams: your PM tool is always current, your workflow always followed, and you get visibility without chasing anyone for status.

Get Running in Three Steps

1 Sign Up and Connect app.flydocs.ai
2 Install the CLI npm install -g @flydocs/cli
3 Start Building /start-session

Quick start

Step 1: Sign up and connect your PM tool

Create your account at app.flydocs.ai, connect Linear or Jira from the portal, and generate an API key. Workspace configuration lives in the portal; the CLI reads it from there. See Sign Up and Connect for the full walkthrough.

Step 2: Install the CLI

Terminal
npm install -g @flydocs/cli && flydocs init

Requires Node.js 18.17+ and Python 3.8+. The init command authenticates with your API key, pulls your workspace configuration, and installs all managed artifacts (hooks, skills, commands, and config files). See Installation for details.

Step 3: Start working

Open your AI coding tool and run:

AI session
/start-session

Your agent loads active issues from Linear or Jira, checks the previous session summary, and picks up where you left off. No pasting context, no searching for the right ticket.


What you get

  • PM integration that runs itself: Linear and Jira stay in sync as you work. Issues transition, comments post, and status updates land automatically through the relay API. No one opens the PM tool to update a ticket.
  • Enforced workflow, not suggested workflow: 7 deterministic hooks across 5 event types handle formatting, context loading, script approval, transition validation, and stop gates. These are standards that run, not instructions the AI can choose to ignore.
  • Context intelligence: workspace topology, session memory, and context graphs mean the agent knows your project, your services, and your recent decisions. Tomorrow's session picks up where today's left off.
  • Skills over MCPs: active skills replace MCP servers at 5-20x less overhead. Loaded on demand, deterministic execution, no background processes. Same capabilities, fraction of the cost.
  • Works everywhere: Cursor, Claude Code, Warp, Codex, and Windsurf. One installation, consistent behavior across tools. The whole lifecycle is structured: capture, refine, activate, implement, review, validate, close.

What's coming

See the roadmap for what we're building next, or join the Discord for support and product updates.


Next