GraphyLoop

The one-command agentic workflow kit for AI coding agents

GraphyLoop is a one-command agentic workflow kit for AI coding agents: it wires Claude Code, OpenCode, Codex, Cursor, Gemini CLI, Oh My Pi and DeepSeek Harness with a 25-agent squad, parallel wave planning, persistent memory and an MCP server. Zero dependencies.

npx graphyloop
Read the README npm package Star on GitHub
  • 7 harnesses
  • 25-agent squad
  • 15 MCP tools
  • 70 skills
  • Zero dependencies
  • Node ≥ 20
  • MIT

Quick start

  1. Run the installer

    npx graphyloop — a bare command installs. It wires every harness it detects, or all seven on a fresh machine.

  2. Restart your harness

    Close and reopen your terminal or editor, then open a real project, not your home directory.

  3. Start the workflow

    Ask your agent to run /chadi-init. Codex: /prompts:chadi-init. Oh My Pi and Cursor: ask it to run the graphyloop workflow init.

Target specific harnesses with npx graphyloop --harness claude,omp. Verify with npx graphyloop doctor — it ends with GRAPH_LOOP_DOCTOR_OK when everything is wired.

Supported harnesses

  • OpenCode

    --harness opencode

    Plugin with graphyloop_* tools, 26 agents, 15 /chadi-* commands.

  • Claude Code

    --harness claude

    MCP server, 26 agents, 15 slash commands, workflow rules.

  • Codex

    --harness codex

    MCP server and 15 prompts (/prompts:chadi-init).

  • Cursor / Windsurf

    --harness cursor

    MCP server plus the workflow as an AGENTS.md rule.

  • DeepSeek Harness

    --harness dsh

    MCP server, 26 role prompts and the graphyloop-squad skill.

  • Oh My Pi

    --harness omp

    MCP server in ~/.omp/agent/mcp.json, 26 squad agents, skills. New in 0.5.0.

  • Gemini CLI

    --harness gemini

    MCP server in ~/.gemini/settings.json, /chadi-* TOML commands and the bundled skills in ~/.gemini/skills/. New in 0.5.0.

Per-harness file locations and verification: harness guide.

How the workflow works

Workflow v2 is harness-neutral: the same rules run in every harness above.

  1. Recall

    Search memory first (memory_search); record the outcome last.

  2. Size

    At most 3 known files is an inline edit. 4+ files or unmapped code is a build.

  3. Context pack + contract

    One batched read pass writes ctx-<slug>.md; the frozen contract-<slug>.md is the only cross-lane dependency.

  4. One wave

    All independent lanes dispatch in one parallel call, one lane per layer, strictly disjoint files.

  5. Verify

    Run the changed path. Tests alone are not proof; UI is checked in a real browser.

  6. Report

    Changed / Files / Verified (verbatim output) / Skills / Blocked.

Features

  • Swarm orchestration

    Spawn, distribute and track agents. State lives in .graphyloop/state.json; no API keys.

  • Persistent memory

    Decisions, patterns and lessons survive restarts and context compaction.

  • 25-agent squad

    Explorer, backend, frontend, test, security, reviewer, architect, data, devops and more.

  • Wave planner

    plan_feature turns a request into contract, parallel builders, integration and parallel verifiers.

  • Universal MCP bridge

    15 tools that work in every MCP-capable harness.

  • 15 slash commands

    /chadi-init, /chadi-plan, /chadi-review, /chadi-deploy and more.

  • 70 bundled skills

    Contract-first waves, API hardening, accessibility, Supabase, Vercel, secrets. Your own skills are never overwritten.

  • Config safety

    Timestamped backups, no overwritten keys, idempotent re-runs, uninstall removes only byte-identical files.

  • Zero dependencies

    Pure Node 20+, MIT licensed, identical on Windows, macOS and Linux.

Harness alone vs + GraphyLoop

CapabilityHarness alone+ GraphyLoop
Agent collaborationIsolated sessionsSwarm with shared memory
OrchestrationManualContext pack, frozen contract, one parallel wave
MemorySession-onlyPersistent, searchable, survives restarts
Multi-harnessOne tool, one rules fileSame workflow in 7 harnesses
SetupHand-written rules per toolnpx graphyloop
Safety—Backup-first merges, content-matched uninstall

Frequently asked questions

What is GraphyLoop?

GraphyLoop is a one-command agentic workflow kit for AI coding agents. It installs a 25-agent squad, parallel wave planning, persistent memory and an MCP server into the harnesses you already use. It is open source (MIT) with zero runtime dependencies.

Does GraphyLoop work with Claude Code, Cursor and Gemini CLI?

Yes. For Claude Code it registers an MCP server in ~/.claude.json, installs 26 agent files and 15 /chadi-* slash commands, and adds the workflow rules. For Cursor and Windsurf it adds the MCP server to ~/.cursor/mcp.json and the rules to ~/.cursor/rules/. For Gemini CLI (since 0.5.0) it adds the MCP server to ~/.gemini/settings.json and installs the commands as TOML files in ~/.gemini/commands/; an existing GEMINI.md is never overwritten.

Which other harnesses are supported?

Seven in total: OpenCode, Claude Code, Codex, Cursor, DeepSeek Harness (dsh), Oh My Pi (omp) and Gemini CLI. Anything else that speaks MCP can use the same server via npx graphyloop mcp. See docs/harnesses.md for what is written where.

Is GraphyLoop free?

Yes. It is MIT licensed and free to use, fork and modify. It uses the model your harness already has; GraphyLoop itself adds no subscription or usage fee.

Does GraphyLoop need API keys?

No. The swarm, memory and MCP tools run locally with no API keys. The only optional key is DEEPSEEK_API_KEY, for letting the engine call DeepSeek directly, and nothing requires it.

How is it different from a plain AGENTS.md or CLAUDE.md?

A rules file is static text that one tool reads. GraphyLoop adds a live engine behind an MCP server (persistent memory, a wave planner that writes the contract and context pack, task and file-ownership tracking) and a 25-agent squad, and installs the same workflow into seven harnesses instead of one.

How do I uninstall GraphyLoop?

Run npx graphyloop uninstall. It removes the core in ~/.graphyloop/, the MCP entries it added, and agents, commands, skills and rules files that are still byte-identical to what it shipped. Anything you edited, and all backups, stay.

Does GraphyLoop send my data anywhere, and does it work on Windows?

It sends nothing anywhere: state and memory are plain files in <project>/.graphyloop/, the MCP server runs locally over stdio, and stored credentials are chmod 600 and never returned to the model. It works on Windows, macOS and Linux; CI runs all three on Node 20, 22 and 24, and on Windows a server-guard keeps npm run dev from hanging an agent session.