archcore-ai/cli

View on GitHub ↗

Git-native context engineering CLI and MCP server for AI coding agents. Keep specs, ADRs, rules, plans, and project knowledge in Git.

60 ★4 forksGoUpdated 18d ago

What you need to know

A git-native context layer for AI coding agents: a Go CLI and local stdio MCP server that keeps specs, architecture decisions, rules, and plans in .archcore/ and serves them to any MCP-compatible agent.

Install

curl -fsSL https://archcore.ai/install.sh | bash (macOS/Linux)
irm https://archcore.ai/install.ps1 | iex (Windows)
go install github.com/archcore-ai/cli@latest

Usage

  • •Run archcore init to scaffold .archcore/ and wire up agent integrations
  • •Ask your agent to record decisions or load relevant ADRs and rules
  • •Use archcore mcp as the local stdio MCP server

Key features

  • ✓19 document types across knowledge, vision, and experience layers
  • ✓4 relation types and 10 MCP tools
  • ✓Hooks for Claude Code, Cursor, Gemini CLI, and GitHub Copilot
  • ✓MCP integration for 8 agents
  • ✓Works in an empty repo via init_project
  • ✓SHA-256 verified self-update

Best for

Teams that want project decisions and conventions versioned in Git and loaded automatically by any coding agent.

Caveats

  • ⚠Not for chat memory or prompt libraries
  • ⚠Requires Go 1.25+ to build from source
Platforms: macOS · Linux · WindowsClients: Claude Code · Cursor · Codex CLI · GitHub Copilot · Gemini CLI · OpenCode · Roo Code · Cline

Reviewed 2026-08-11

Topics

adrai-agentsai-codingarchitectureclaude-codecoding-agentscontext-engineeringcursordeveloper-toolsgit-nativeharnessharness-engineeringmcpmcp-servermodel-context-protocolproject-contextspec-driven-developmentspecifications
Stars
60★
Forks
4
Language
Go
License
Apache-2.0
Created
2026-02-18
Last push
2026-08-17