DOCSEARCHbyAlgolia
Submit Docs

Okto Pulse

website·Indexed 30 days ago

https://pulse.oktolabs.ai/

A local-first SDLC workbench for teams building software with AI coding agents, designed to maintain product intent, specifications, and tasks.

Tokens
935
Snippets
2
Records
8
Agent score
0%

What's inside Okto Pulse

  1. Overview of Okto Pulse

    Okto Pulse is a local-first SDLC workbench for teams building software with AI coding agents. It ensures that product intent, specifications, tasks, tests, and validation evidence remain connected and inspectable. Unlike Okto Nexus (which coordinates agent handoffs) or Okto Neuron (which provides cited graph memory), Pulse governs what must be built and how delivery is validated. It runs locally, requires no account, and is compatible with MCP (Model Context Protocol) coding agents.

  2. Manage project memory with the Pulse Knowledge Graph

    The Knowledge Graph in Okto Pulse serves as durable project memory, storing typed connections between initiative records (Ideations, Refinements, Specifications, Tasks, Tests, and Bugs). Unlike LLM-generated conversational answers, Pulse retrieves connected graph records to provide provenance for decisions, constraints, and rejected alternatives. This allows agents to query why an implementation exists, what prior learnings apply, and which recorded bug violated a specific constraint.

  3. Understand Okto Pulse product positioning and constraints

    Okto Pulse is a local-first SDLC workbench for teams building software with AI coding agents. It is designed to make delivery inspectable against requirements and recorded evidence, ensuring captured and consolidated decisions remain queryable with provenance.

    Key distinctions and constraints:

    • Not a coordination bus: Pulse is not a multi-agent coordination bus; that role is handled by Okto Nexus.
    • Knowledge Graph behavior: The Knowledge Graph does not call an LLM or produce ungrounded chat answers; it retrieves structured sources for a connected agent to interpret.
    • Capabilities: Pulse does not autonomously discover every ambiguity or automatically consolidate every decision into the Knowledge Graph.
    • Licensing: The product is source-available under the Elastic License 2.0 (not open source).

  4. Install and start Okto Pulse

    Okto Pulse requires Python 3.11 or newer. It runs locally and does not require an account. The installation process involves installing the package, initializing the local environment (which creates local data, a default board, and a .mcp.json file), and starting the server.
    pip install okto-pulse
okto-pulse init
okto-pulse serve

  5. Follow the Canonical Delivery Workflow in Pulse

    The standard delivery lineage in Okto Pulse follows this sequence: Stories → Ideation → Refinement → Spec → Sprint → Tasks / Tests / Bugs.

    1. Outcome Definition: A human describes the desired product outcome.
    2. Scope Evaluation: An agent uses structured questions to clarify ambiguity, risk, and constraints.
    3. Specification: Stable context is turned into a specification (requirements, acceptance criteria, architecture, etc.).
    4. Execution: The specification is divided into a sprint with traceable tasks, tests, and bugs for agents to execute via MCP.
    5. Governance: Gates and validators inspect readiness, coverage, and drift.
    6. Consolidation: Stable decisions are moved into the Knowledge Graph for future retrieval.

  6. Distinguish between Okto Pulse, Nexus, and Neuron

    The Okto ecosystem consists of three primary components:

    • Okto Pulse: A local-first SDLC workbench that manages product intent, specifications, and tasks to keep delivery aligned with requirements.
    • Okto Nexus: The coordination layer responsible for agent communication, ownership, handoffs, approvals, and audit history.
    • Okto Neuron: A local-first knowledge graph built from team notes that provides answers with citations to their sources.

  7. Okto Pulse default network configuration

    By default, Okto Pulse exposes the following local endpoints:

    • Local UI and API: Port 8100
    • MCP endpoint: Port 8101 at /mcp

  8. Configure and reset MCP agents in Okto Pulse

    The okto-pulse init command generates the initial .mcp.json configuration. If the MCP agent configuration needs to be regenerated (for example, after changing the server port), use the --agents flag.
    okto-pulse init --agents