OktoLabs Documentation

website·Indexed 29 days ago

https://oktolabs.ai

Tools for teams delivering software with AI agents, focusing on maintaining requirements, context, control, ownership, and evidence.

Tokens
722
Snippets
3
Records
6
Agent score
6%

What's inside OktoLabs

  1. Overview of OktoLabs Product Suite

    OktoLabs provides a set of tools designed for teams delivering software with AI agents while maintaining requirements, context, and evidence. The suite consists of:

    • Okto Pulse: A local-first delivery workspace that organizes work using Specification-Driven Development (SDD) and Test-Driven Development (TDD). It connects intent, requirements, tasks, and validation evidence.
    • Okto Nexus: A coordination and communication layer for AI agents and tools to manage ownership, handoffs, and human approvals.
    • Okto Neuron: A local-first knowledge graph built from Markdown notes that provides cited answers via an MCP server.
    • Okto Grafx: An embedded, local-first graph database for Python that serves as the substrate for Pulse and Neuron.
  2. Install and use Okto Neuron

    Okto Neuron creates a local-first knowledge graph from a team's Markdown notes, allowing agents to query shared memory with citations via an MCP server. It is not available on PyPI and must be installed via script.
    # macOS or Linux
    curl -fsSL https://raw.githubusercontent.com/OktoLabsAI/okto-neuron/main/install.sh | bash
    
    # Windows (PowerShell)
    powershell -NoProfile -ExecutionPolicy Bypass -Command "irm https://raw.githubusercontent.com/OktoLabsAI/okto-neuron/main/install.ps1 | iex"
  3. Install and use Okto Nexus

    Okto Nexus provides a governed coordination layer for AI agents (e.g., Claude, Codex, Cursor) and tools. It is used to manage agent handoffs, assign explicit ownership to tasks, and require human approval for sensitive actions without manually copying context.
    pip install "okto-nexus[serve]"
  4. Install and use Okto Pulse

    Okto Pulse is used to preserve business reasons behind code changes, turn product intent into agent-ready tasks, and connect acceptance criteria to validation evidence. It is a local-first workspace for software built with AI.
    pip install okto-pulse
  5. Coordinate AI agents using Pulse and Nexus

    For an integrated delivery flow, combine Pulse (for structure/verification) and Nexus (for coordination):

    1. Human describes need $ ightarrow$ Pulse structures requirements/tasks.
    2. Nexus transfers ownership to an implementation agent.
    3. Agent executes $ ightarrow$ records status/evidence in Pulse.
    4. Nexus hands result to an independent reviewer.
    5. Reviewer validates criteria $ ightarrow$ submits findings to Pulse.
    6. Nexus closes handoff and notifies agents.
    7. Pulse retains the verified delivery record.
  6. Implement the ADLC operating model

    The AI Delivery Life Cycle (ADLC) is the operating model powering OktoLabs tools. It consists of six stages:

    1. Ideate: Capture business reason and desired outcome.
    2. Refine: Resolve ambiguity, constraints, and decisions.
    3. Specify: Record requirements and acceptance criteria.
    4. Execute: Coordinate agents while linking tasks to specifications.
    5. Validate: Compare delivery against tests, criteria, and evidence.
    6. Learn: Preserve verified knowledge in a cited graph (using Okto Neuron).