Getting Started with Okto Pulse

Getting Started

Okto Pulse is the first source-available platform for the Agentic Development Life Cycle (ADLC) — a local-first, spec-driven project board with native MCP (Model Context Protocol) support. Humans and AI agents plan, refine, and ship on the same board without stepping on each other.

Prerequisites

• Python 3.10+
• pip (or pipx / uv if you prefer)
• An MCP-aware agent (Claude Code, Cursor, Windsurf, or any MCP client) — optional, but recommended.

Install

pip install okto-pulse

No Docker, no accounts, no cloud. Pulse runs from SQLite on your machine.

Initialize

okto-pulse init

This creates a local data directory and generates a starter .mcp.json you can point your agent at.

Run the board + MCP server

okto-pulse --api-port 8100 --mcp-port 8101 serve

• Web UI: http://localhost:8100
• MCP endpoint: http://127.0.0.1:8101/mcp

Point an agent at your board

Drop the generated .mcp.json next to your agent’s config:

{
  "mcpServers": {
    "okto-pulse": {
      "url": "http://127.0.0.1:8101/mcp?api_key=dash_xxx"
    }
  }
}

Your agent now has access to 150+ tools spanning ideations, refinements, specs, sprints, cards, validations, and analytics.

What next

• The Pulse pipeline — how ideations flow through to shipped tasks.
• Agent roles — the five built-in permission presets.
• MCP integration — wiring Claude Code, Cursor, and others.

Or jump straight into the full landing at pulse.oktolabs.ai.