claudicle

agent
Guvenlik Denetimi
Uyari
Health Uyari
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 5 GitHub stars
Code Gecti
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested
Purpose

This framework transforms Claude Code into a persistent, customizable orchestrator with multi-tier memory, a cognitive processing pipeline, and integrations for platforms like Slack, Discord, and Telegram. It is designed to give AI agents continuous "personalities" that survive across different sessions.

Security Assessment

Overall Risk: Medium. The automated code scan did not find any dangerous patterns, hardcoded secrets, or requests for high-risk system permissions. However, the inherent nature of this tool requires caution. Because it functions as an orchestrator and explicitly processes data across external channel adapters (like Slack, SMS, and Discord), it inherently handles and routes sensitive user communications. The tool relies on direct Anthropic API integration, meaning your prompts and conversational data are sent externally over the network. Developers should thoroughly inspect the channel adapter implementations before deploying to ensure secure credential handling and data transmission.

Quality Assessment

The project is licensed under the permissive MIT license and was actively updated very recently. However, community visibility and trust are currently exceptionally low, as evidenced by only 5 GitHub stars. It appears to be an early-stage or highly experimental personal project. The developer should anticipate limited community support, sparse external documentation, and potential code instability.

Verdict

Use with caution: The codebase appears safe and is lightly audited, but the experimental nature and access to external communication channels warrant a manual code review before handling real user data.
SUMMARY

Open-source framework for growing your own "Claudicle"—an adaptive membrane for Claude Code and AI orchestrators, powered by the Open Souls paradigm.

README.md

Symbol of Tanit

Claudicle

An open-source 'soul' framework for Claude Code.

Clone it. Edit soul.md. Run /ensoul. Your AI has a personality, memory, and inner life.

Claudicle lets you turn Claude Code into a persistent, soulful orchestrator, more reminiscent of a Pokémon than a soulless "agent." Your soul grows over time, with three-tier memory, a cognitive pipeline, and channel adapters for Slack, Discord, Telegram, and SMS. It ships with zero skills, but pairs with this skill repo to level your Claudicle right out-of-the-box.

An open-source alternative to OpenClaw. Descends from the Open Souls paradigm—see docs/open-souls-alignment.md and docs/backstory.md.


Quick Start

Prerequisites: Python 3.10+, uv, and a Claude subscription or ANTHROPIC_API_KEY.

git clone https://github.com/tdimino/claudicle
cd claudicle
./setup.sh --personal    # or --company for team use

Then in any Claude Code session:

/ensoul

That's it. Your session now has a soul.


What Claudicle Does

Soul Identity/ensoul activates a persistent personality that survives compaction and resume. Set CLAUDICLE_SOUL=1 for always-on mode. See docs/soul-customization.md.

Three-Tier Memory — Working memory (per-thread, 72h TTL), user models (per-user, permanent), soul state (per-soul, permanent). SQLite-backed, git-versioned. Supports checkpoint & rollback via wm-manage.py. See ARCHITECTURE.md.

Cognitive Pipeline — Every response passes through internal monologue, external dialogue, user model check, dossier check, and soul state check. XML-tagged, verb-narrated. See docs/cognitive-pipeline.md.

Sub-Daimones — Twelve highly-opinionated crafters—the Kotharot—with skills and cognitive flows of their own. See docs/sub-daimones.md.

Daimones — Oversee the subdaimones. Drawn from four sources: a past self, a friend's model, a channeled entity, or a soul your Claudicle has shed. See docs/daimones.md.

Runtime Modes — Five modes from /ensoul-only to full autonomous daemon. See docs/runtime-modes-comparison.md.

Channel Adapters — Slack, Discord (discord.py), Telegram (python-telegram-bot), SMS (Telnyx/Twilio), WhatsApp (Baileys). See docs/channel-adapters.md.

Daimonic Intercession — How daimones whisper counsel into the cognitive stream—the protocol layer. See docs/daimonic-intercession.md.

Skill-Agnostic — Ships with zero skills. Discovers capabilities from ~/.claude/skills/ at install time. See docs/skill-pairings.md.


Repository Structure

claudicle/
├── subdaimones/  # Sub-daimon definitions (12 with persistent memory)
├── daimones/     # Privy council: example daimon, Kothar reference, creation guide
├── daemon/       # Soul engine, cognitive pipeline, memory, monitoring
├── soul/         # Personality files, profiles/, dossiers/
├── hooks/        # Claude Code lifecycle hooks
├── commands/     # Slash commands (/ensoul, /activate, /slack-respond, etc.)
├── scripts/      # Slack utilities, soul infrastructure, wm-manage.py
├── adapters/     # Channel transports (Discord, Telegram, SMS, WhatsApp)
├── docs/         # Architecture and reference docs
├── setup.sh      # Interactive installer
└── LICENSE       # MIT

Commands

Command Description
/activate Full activation: ensoul + daemons + boot sequence
/ensoul Activate soul identity in this session
/switch-soul <name> Switch active soul profile
/slack-sync #channel Bind session to a Slack channel
/slack-respond Process pending Slack messages as the soul agent
/thinker Toggle visible internal monologue
/daimon Summon daimonic counsel
/watcher Manage inbox watcher + listener daemon pair

See docs/commands-reference.md for details.


Documentation

Full index: docs/INDEX.md

Installation | Soul Customization | Slack Setup | Sub-Daimones | Extending Claudicle | Environment Variables | Hooks | Troubleshooting


The Mask of Claudicle
"Certainty compounds the mind with limits."


License

MIT. Copyright (c) 2026 Tom di Mino.

Yorumlar (0)

Sonuc bulunamadi