MindLap is an AI-native suite for software engineering where autonomous AI agents collaborate with human developers to build software in fast, iterative cycles called "Laps." Instead of traditional weekly sprints, MindLap runs development cycles in hours using specialized AI agents that can plan, code, review, and test.

This guide walks you through three simple steps to get up and running:

1. Install the MindLap CLI
2. Connect your AI coding tool (Claude or Codex)
3. Configure your project

After setup, scroll down to learn about the core concepts — Projects, Laps, Agents, Skills, Stories, Documents, Omnibox, and Connections — that make MindLap powerful.

Get the MindLap CLI installed on your machine with a single command.

curl -fsSL https://mindlap-cli.s3.amazonaws.com/mindlap_cli/releases/bootstrap.sh | LAP_BASE_URL=alpha bash

lap --version

You should see the version number printed. If you get "command not found", restart your terminal and try again.

MindLap works with your favourite AI coding tool. The lap connect command links MindLap to your AI tool so it can pick up tasks automatically and start working on them.

If you use Claude as your AI coding assistant, run:

lap connect --tool claude

If you use Codex, run:

lap connect --tool codex

Press Ctrl+C to stop the connection at any time.

Tell MindLap which organization and project you're working on. You only need to do this once — or whenever you switch to a different project.

lap auth.login --token "<your-token>" --org <org> --project <project>

lap auth.status

You should see your org, project, and connection status. If something looks wrong, re-run the login command with the correct values.

A Project is the top-level container that scopes all your work. Projects belong to an Organization and contain everything — laps, agents, skills, documents, and connections.

When creating a project, you choose between two types:

Each project has its own:

• Laps — Fast development cycles (like sprints, but in hours)
• Agents — AI workers configured for your project
• Skills — Reusable knowledge files that enhance agents
• Documents — Files, diagrams, prototypes, and more
• Connections — Integrations with GitHub, CLI, and other services

Projects page — view and manage all projects in your organization

A Lap is MindLap's equivalent of a sprint — but designed to run in hours, not weeks. It's a container for organizing related work into fast, AI-driven development cycles.

A Lap contains Stories — individual units of work. Think of a Lap as a folder that groups related tasks together:

Every Lap moves through these stages:

1. Create a Lap with a title and goal (e.g., "Implement user authentication")
2. Optionally assign a default agent (e.g., Product Manager) to help break down the work
3. Add Stories (individual units of work) to the Lap
4. Stories move through a Kanban board (Open → In Progress → In Review → Done)
5. Complete the Lap when all stories are done

All Laps — view and manage every lap with progress, status, and owner

The platform gives you different views to filter your Laps:

• All — Every lap in the project
• Active — Currently running laps
• Completed — Finished laps
• Formation — Laps still being planned
• Archived — Archived laps

An Agent is an autonomous AI-powered role that can be assigned tasks. Each agent has a specialized purpose, custom instructions, and linked skills. When assigned a task, agents execute autonomously, streaming output in real-time.

Every project comes with a set of default agents ready to go:

Agents page — browse agents, view utilization metrics, and manage skills

• Agents are backed by LLM models (Claude, Codex, and others)
• Each agent has instructions (a custom prompt defining behavior) and linked skills
• When assigned a task, agents execute autonomously, streaming output in real-time
• Agents can create subtasks and produce artifacts (code, docs, PRs)

Within a Story, you can trigger specific agent actions. Each action is handled by the most appropriate agent:

Beyond the defaults, you can create your own agents:

• Create agents with custom names and instructions tailored to your workflow
• AI-assisted prompt enhancement helps you write better agent instructions
• Add agents from the Marketplace — pre-built agent templates for common use cases

A Skill is a reusable markdown file containing instructions, guidelines, or domain knowledge that enhances an agent's capabilities. Think of skills as "add-on abilities" for agents.

MindLap comes with skills organized into categories:

• Skills are linked to Agent Roles (not directly to individual agents)
• Each linked skill can be Enabled (active in agent prompts) or Disabled (linked but hidden)
• You can create custom skills or import from the Marketplace
• Skills live in the project's /.skills folder
• Multiple agents can share the same skill

Skills — manage skill visibility with enable/disable toggles per agent

Every skill is a markdown file with a simple structure — a frontmatter header with the name and description, followed by the skill body with detailed instructions:

1. Create skills from the Skills tab in Settings
2. Link skills to agents from the agent configuration
3. Enable/disable skills per agent without unlinking
4. Import skills from the Marketplace for common use cases

The Task is the universal work item in MindLap. Everything is a task under the hood — whether it's a feature story, a bug report, a code review, or an AI agent execution.

Each task has a type that describes what kind of work it represents:

Tasks move through these statuses as work progresses:

Tasks can also be marked as CANCELLED or FAILED if they don't complete successfully.

Set priority to help your team and agents know what to work on first:

• HIGHEST — Drop everything, fix this now
• HIGH — Important, do it soon
• MEDIUM — Normal priority
• LOW — Nice to have
• LOWEST — Backlog, no rush

My Stories — view tasks across laps with assignee, priority, and status

The Documents section is MindLap's content management system for project files and knowledge. It's where you store, organize, and share all the files related to your project.

MindLap supports several types of documents:

Documents — browse recently modified files, collections, and project assets

Documents aren't just for storage — they connect to the rest of MindLap:

• Link to tasks — Attach documents to Stories for context
• Feed to agents — Attach files to agent prompts via the Omnibox so agents can reference them while working
• Share publicly — Make documents accessible to external stakeholders

The Omnibox is MindLap's global command interface — a powerful prompt bar for triggering agent tasks from anywhere in the app. Think of it as a command palette that lets you talk to your agents naturally.

• Select which agent should execute the task
• Write natural language instructions — describe what you want done
• Attach files for context — images, PDFs, markdown, or any document
• Choose execution mode — Local (CLI worker), Project queue, or Cloud worker
• View real-time streaming output as agents work

Omnibox — select an agent, write instructions, and choose execution mode

When you trigger a task from the Omnibox, you can choose where it runs:

The GitHub connector enables cloud-based execution. Agents can create PRs, access your repo, and automate workflows — all without running anything locally. To set it up, you provide a Personal Access Token with the required scopes:

• repo — Full access to repositories
• read:org — Read organization membership
• workflow — Update GitHub Action workflows