OpenThunder Docs

Short, truthful, reference-grade documentation for OpenThunder: the AI developer environment where you build with AI, verify what it produces, and own the outcome. Everything here describes what ships today, nothing more.

OpenThunder runs on four surfaces , Desktop, for VS Code, Cloud, and CLI , over one engine. The understand, risk, and verification work is deterministic and local-first: it runs with no AI key and never uploads your source by default.

Start here (tutorials)

Getting started: install OpenThunder Desktop and get first value in under 5 minutes, offline, no account.
Zero-install diff review: paste a git diff and get an instant change review. Runs 100% in your browser; the diff never leaves your machine.

Guides (how to)

Task-focused walkthroughs for the work you actually do:

Verify an AI-generated change before you commit: the loop to run after your agent edits code.
Decide if a change is safe to ship: read a SHIP / CAUTION / HOLD verdict, blast radius, and the Evidence Pack.
Understand a repo you didn't write: get oriented in an unfamiliar or AI-generated codebase, from the real code.
Set up a verified mission for high-risk work: auth, billing, DB migrations, and security changes, with an independent reviewer.
Connect your AI agent (MCP): wire Claude Code, Codex, Cursor, or Copilot into your repo graph, keyless and local.
Use OpenThunder in CI: gate merges on the same verdict as your local commits.
Get a repo development-ready: openthunder ready detects the stack and prepares the environment.
Check security, including AI-era risks: classic + AI security (OWASP LLM), blast-radius triage, and the security gate.
Check dependencies and supply chain: SBOM, provenance, and an honestly-labeled trust report.
Roll out OpenThunder to your team: local-first, CI gates, and privacy-first shared visibility.

Concepts (understand)

Core concepts: Missions, the Lenses, Change Verification, AI Context Packs, Blast Radius, and the MCP Context Brain.
Evidence Packs and provenance: why OpenThunder produces evidence, not just an AI summary , the differentiator, and how every conclusion is labeled by source and confidence.

Reference

CLI reference: every openthunder command, generated from the real --help output.
MCP tools, resources, and prompts: the full agent-facing surface.

Surfaces

One engine, four surfaces:

OpenThunder Desktop: the free local app for macOS, Windows, and Linux.
OpenThunder for VS Code: understand, verify, and connect your agent inside the editor.
OpenThunder Cloud: the optional, privacy-first team control plane.
CLI: bundled with Desktop; see the CLI reference.

Trust and privacy

Trust Center: exactly what runs locally, what leaves only when you choose, and the three execution modes.
Privacy FAQ: what never leaves your machine, and what the Cloud is for.

OpenThunder emits verification evidence. Skills Tech Talk turns it into practice and Repo Mastery learning paths. Those are separate products; they are linked from OpenThunder, not built into it.