LoopBoard is the Seii.top flagship product: an AI-native product, project and QA management system for teams where coding agents maintain structured records and humans review the evidence.
Release readiness summary
Progress is calculated from real work package fields, review state falls back conservatively, and every change can be traced to a prompt, CLI command, HTTP route, or Markdown import.
The homepage now presents Seii.top as a LoopBoard-first company: project truth at the center, compliant document transformation beside it, and CaptureW retained as the auxiliary web capture and knowledge portal.
The main Seii.top product: AI coding agent-native product, project and QA management. Agents maintain structured records; humans steer, approve and inspect the audit trail.
A document compliance engine that turns internal materials, requirements, tests and delivery notes into safer professional artifacts for review and external delivery.
The retained web capture and knowledge portal: browse captures, manage Books and wiki knowledge bases, and keep owned knowledge available beside LoopBoard work.
Traditional PM tools assume humans type every field. LoopBoard assumes humans provide raw evidence, agents structure it, and the system must keep every claim traceable.
Requirements, plans and bugs are primarily written by AI coding agents from real human-provided material, then reviewed by product, QA and project owners.
Every manageable entity needs command-line and HTTP paths backed by the same business functions, so agents can operate without clicking through a browser.
SQLite remains authoritative while Markdown mirrors give agents and humans a readable working surface for requirements, bugs, work packages, sprints and meetings.
Progress bars, review status and dashboards must trace back to real persisted fields. When evidence is missing, the product falls back conservatively.
Managed entities use soft archive semantics so requirements, plans, bugs and meeting notes remain auditable and recoverable across the project lifecycle.
Agent judgments that are real but not yet commitments live in a separate structured space, preserving exploration without polluting formal scope.
A companion product turns internal materials into professional deliverables while scanning for identity, customer, architecture, strategy and evidence clues.
CaptureX and CaptureW stay in the portfolio as the owned knowledge capture layer: multimodal intake, Books, wiki knowledge bases and Git-backed storage.
LoopBoard is designed around the real workflow: humans provide messy context, agents structure it, and the system preserves the evidence trail.
Meeting notes, screenshots, defect descriptions and external documents enter through an agent, CLI, HTTP route or Markdown working copy.
Requirements, work packages, bugs, sprints and notes are normalized into fields, links, status and audit-friendly Markdown.
Dashboards derive progress from explicit persisted fields. Unknown review states stay pending until the evidence says otherwise.
Soft archive, prompt logs, LLM output archives and report snapshots make project decisions recoverable instead of silently rewritten.
The product is not a Jira clone with an AI assistant bolted on. It is built around the fact that coding agents read repositories, run commands, edit Markdown, call APIs and produce most structured project records from raw human material.
Document Compliance Transformation scans for explicit and implicit clues, preserves the original logic, rewrites the structure, and produces professional documents that can be reviewed, reused, trained on, or shared safely.
Clue scan — identity, customer, architecture, strategy and evidence risksLogic extraction — requirements, fields, flows, tests and acceptance criteriaProfessional rewrite — clear sections, stable terms and delivery-grade languageQuality gate — negative search, structure checks and fact-preservation reviewInventory
Risk mapping
Transformation
Verification