AI Decision System
Replayable, offline-verifiable records for AI-assisted go/no-go decisions: experiment promotions, typed model answers and human-approved actions.
Your coding agent turns the notes and decisions you already have into a company brain, and you approve what goes in. Codex and Claude Code answer from it and show the source behind each answer. Plain
Developer toolsPython · TypeScriptGitHub @MarvinAmador7OnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
Owners of public repositories under a personal GitHub account can sign in to verify ownership and manage the introduction themselves. No administrator approval is needed.
Sign in, generate a verification file, and commit it to this repository’s default branch within 30 minutes. Files on other branches or forks do not count. Then check the file below to manage the project without administrator approval. Keep the file to retain your access.
If needed, you will go to sign-in. After signing in, return here and press your chosen verification button again.
Run /nomorevibe verify in your project folder to verify ownership and manage the introduction yourself.
Verify in one minuteAsk about other verification options
Your coding agent turns the notes and decisions you already have into a company brain, and you approve what goes in. Codex and Claude Code answer from it and show the source behind each answer. Plain files in a Git repo you own.
Taken from the repository README
Loading the full README.
Read the full README ↗Company answers, with evidence you can inspect.
Groundray helps your team and its AI agents work from a portable company brain: decisions, procedures, definitions, and original records kept in Markdown and Git. Ask a business question, inspect the evidence, and propose updates as your company learns.
Website · Get started · How teams use it · Documentation · Releases
What can we tell this customer about price and launch? Which numbers belong in our review, and what still needs a decision?
Groundray's workflow is answer → evidence → guidance:
Use it to recover the reasoning behind a decision, find the current procedure, reconcile conflicting figures, or prepare a customer response with its supporting records. When the evidence is incomplete, keep that gap visible.
Install the plugin on macOS or Linux, Intel or ARM. You need Git and an installed Codex CLI; the package includes Groundray's CLI and keyword search engine. No separate Python, Node, QMD, or model download is required.
curl -fsSL https://github.com/MarvinAmador7/groundray/releases/latest/download/install-groundray | sh
Start a fresh Codex session using the executable selected by the installer. Connect an existing Groundray brain, replacing the URL below with your company's brain repository:
Use $groundray:groundray-setup to connect my company brain from <repository-url>.
Then ask a question:
Use $groundray:groundray. What can we promise this customer about price and
launch? Show the evidence and what still needs a decision.
Use the same native package with an installed Claude Code CLI:
curl -fsSL https://github.com/MarvinAmador7/groundray/releases/latest/download/install-groundray | sh -s -- --host claude
Start a fresh Claude Code session, then connect your brain and ask:
/groundray:groundray-setup Connect my company brain from <repository-url>.
/groundray:groundray What can we promise this customer about price and launch?
Show the evidence and what still needs a decision.
A company brain is a separate Git repository. To create your first one, use the CLI below. See the plugin guide for choosing a host executable, upgrading, and removing the plugin.
Install the standalone CLI:
curl -fsSL https://github.com/MarvinAmador7/groundray/releases/latest/download/install | sh
export PATH="$HOME/.local/bin:$PATH"
groundray --version
Create a brain and capture your first draft through the guided workflow:
groundray setup
groundray brain init ./team-brain --name "Team" --slug team --no-publish
groundray knowledge create --brain team
For example, name the entry Customer onboarding and supply its supporting sources when prompted. Index the saved knowledge, then search it:
groundray search index team
groundray search query "customer onboarding" --brain team --json
Setup can also install the standalone agent workflow for Codex, Claude Code, or both. See the CLI guide for connecting an existing brain, reading original sources, and authoring changes.
Brains stay in your Git repositories. Search runs locally, and indexes are rebuildable. Publishing and syncing are explicit; repository access uses your Git host's permissions. Business approval remains a company decision.
See reviewed collaboration and the authoring handbook for the complete workflow.
Early release: 0.3.1. The Codex and Claude Code plugins, standalone CLI, original-source retrieval, and proposal/review/publish/sync workflows are implemented. Native packages and installers are tested on all four supported platform combinations. Real employee handoffs and the human evidence-review experience remain pilot work; see the pilot record.
Retrieval currently uses local keyword search. New setups include the builtin engine; existing QMD configurations are supported. The roadmap covers future capabilities.
Upgrading from Portable KB? See the 0.3.0 migration notes for renamed commands and configuration paths.
| Start here | Details |
|---|---|
| Plugin guide | Codex and Claude Code installation, company setup, upgrades, and removal |
| CLI guide | Brains, search, source capture, edits, sharing, and diagnostics |
| Release guide | Downloads, checksums, version pins, and migration |
| Product direction | Product contract, milestones, and acceptance criteria |
| Evaluations | Reproducible agent trials, evidence, scoring, and related benchmarks |
| Development | Contributor setup, native builds, tests, and architecture |
| Website | React/Vite landing page, local preview, and static build |