Skip to main content

CycleGen — 3次元記憶システム(スキル・記憶ストア)を備えた人間-AI協働フレームワーク

Project description

CycleGen Core

A human–AI collaboration framework with a 3-dimensional memory store (Skill & Memory Store), built on the "Deep Out" design philosophy.

CycleGen turns AI collaboration into a repeatable one-hour cycle: the AI works autonomously, you review and decide, and the context you accumulate is structured and re-injected — so the same model gets qualitatively better at your work over time. CycleGen Core is the open-source (Apache-2.0) personal edition.

"Rent the model. Own the context."

日本語版は README.ja.md を参照してください。


What's inside

CycleGen Core ships an MCP server exposing 19 tools for a semantic, self-ranking memory store:

  • Semantic memory search — recall the right memories by meaning, ranked across 3 axes (Layer / Priority / Context), following Miller's 7±2 to avoid overload.
  • Store / update / pin / archive — capture knowledge as you work; priority rises with use and decays without it.
  • Cycle lifecyclecycle_complete records a work cycle and surfaces promotion candidates.
  • CycleGen Finish (optional docx extra) — convert Markdown to styled .docx.

Install

CycleGen Core runs as an MCP server. The recommended way is via uv:

# Run the MCP server directly (installs on first use)
uvx --from "cyclegen[semantic,docx]" cyclegen-mcp

Or with pip:

pip install "cyclegen[semantic,docx]"
cyclegen-mcp
  • semantic — embeddings backend for memory search (recommended; first run downloads a small model).
  • docx — enables the document_finish / list_finish_templates tools.

Use as a plugin

CycleGen ships a plugin for Claude Code and Codex that wires the MCP server and the cycle discipline (hooks) automatically. See the guide at https://cyclegen.io.

MCP client configuration (manual)

{
  "mcpServers": {
    "cyclegen": {
      "type": "stdio",
      "command": "uvx",
      "args": ["--from", "cyclegen[semantic,docx]", "cyclegen-mcp"]
    }
  }
}

Documentation

License

Apache License 2.0 — see LICENSE. Copyright 2026 rashiku Corp.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

cyclegen-0.1.0.tar.gz (153.8 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

cyclegen-0.1.0-py3-none-any.whl (106.6 kB view details)

Uploaded Python 3

File details

Details for the file cyclegen-0.1.0.tar.gz.

File metadata

  • Download URL: cyclegen-0.1.0.tar.gz
  • Upload date:
  • Size: 153.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.11.15

File hashes

Hashes for cyclegen-0.1.0.tar.gz
Algorithm Hash digest
SHA256 f2a7cd7b094559dcaeebb441a4402b3da07328d979e228067fd2458d2d682065
MD5 11471449ef4370caff80b19860dfd111
BLAKE2b-256 bfe754da6255e4f80f9156246d880de5c929548e6051aab826cb5b5058272362

See more details on using hashes here.

File details

Details for the file cyclegen-0.1.0-py3-none-any.whl.

File metadata

  • Download URL: cyclegen-0.1.0-py3-none-any.whl
  • Upload date:
  • Size: 106.6 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.11.15

File hashes

Hashes for cyclegen-0.1.0-py3-none-any.whl
Algorithm Hash digest
SHA256 c77a656d6069af8e445e908d6ad2a6c408f803f9a0f62fc48b07fc8179b68470
MD5 47ccb7fda6d76ace146d8f006c17b200
BLAKE2b-256 6798b8fd276fb240d50de31069d55ce72e1db0e85b6d094ad455c55c6d8867f1

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page