Skip to content

Repository files navigation

Msingi

Context engineering infrastructure for AI agent sessions.

Msingi is Swahili for foundation — the groundwork you lay before building.

License: MIT Runtime: Go Platform: Cross-Platform Version

Built in Accra. Designed for everywhere.


The Problem

The real cost of AI coding tools is the tokens burned on unstructured exploration. Every cold session costs tokens that should go to creation.

Msingi generates the context foundation in 60 seconds.

Modern Architecture (Go Migration)

Msingi has evolved from a collection of scripts into a high-performance, data-driven engine built in Go.

graph TD
    subgraph "Input Layer"
        JSON["agents.json & skills.json"]
        Tmpl["Markdown Templates"]
    end

    subgraph "Core Engine (Go)"
        Eng["Template Engine"]
        TUI["Bubble Tea TUI"]
        Gen["Project Generator"]
    end

    subgraph "Artifacts"
        Scaffold["Full Project Scaffold"]
        Audit["BOOTSTRAP.md Audit Log"]
    end

    JSON --> TUI
    Tmpl --> Eng
    TUI --> Gen
    Eng --> Gen
    Gen --> Scaffold
    Gen --> Audit
Loading
  • Go (Canonical): Handles structured JSON parsing, metadata injection, and high-fidelity TUI rendering. This is the recommended entry point for all users.
  • Bash (Lightweight): Preserves zero-dependency, offline-first portability. Used as a lightweight stub for minimal scaffolding in constrained environments.

TUI Features

The Msingi interactive wizard provides a premium onboarding experience:

  • Guided Setup: 7 screens covering Project Mode, Type, Details, Intake, Agents, and Skills.
  • Context Pressure Preview: Live estimation of token usage for your current selections (Toggle with p).
  • Skill Inference: Real-time matching of project metadata against the skill registry to suggest relevant capabilities.
  • Responsive Design: Auto-adjusting layouts with sidebars, progress indicators, and paginated lists.

Installation

From Source (Go)

git clone https://github.com/xdagee/msingi
cd msingi
go build -o msingi
./msingi

Windows (PowerShell 7)

.\install.ps1
bootstrap

macOS / Linux (Lightweight Bash)

chmod +x msingi.sh
sudo ln -s "$(pwd)/msingi.sh" /usr/local/bin/msingi
msingi

Developer Ergonomics

  • --dry-run: Preview all file actions without touching the disk.
  • --verbose: Enable detailed logging and template validation.
  • --template-dir: Hot-reload local templates during development.

For AI Agents

If you are an AI agent working on this repository, please start by reading AGENTS.md.


Msingi — the foundation you lay before you build.

Built in Accra. Designed for everywhere.

About

Context engineering infrastructure for AI agents.

Resources

Contributing

Stars

3 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages