Continuous-Claude-v3

Continuous Claude v3 provides a persistent, learning multi-agent environment built around Claude Code. It orchestrates a suite of skills, workflows, and agents to maintain context across sessions, analyze code, and automate complex tasks with memory and continuity. The MCP integration helps you launch and manage the server, expose its capabilities to clients, and coordinate sub-sessions (skills, hooks, and agents) to tackle tasks like debugging, code analysis, and feature planning. Use the included tools to perform TLDR-style code analysis, automate common workflows, and leverage the memory system to recall learnings from prior sessions. The system emphasizes compounding knowledge: it extracts learnings from interactions, stores them, and uses them to inform future actions without losing important context when you resume later.

To interact with the server you’ll typically either run the MCP server via your preferred package manager (uv in this README’s workflow) or connect a client that can issue skill activations, agent prompts, and handoffs. The Skill Activation System automatically injects context to decide which skills and agents are relevant for a given request. You can request tasks like debugging a file, exploring a codebase, or performing a premortem analysis, and the system will propose a curated set of skills (e.g., create_handoff, fix, deb ug) and agents (e.g., scout, debugger) to execute the workflow. This makes it easier to describe intent naturally while still enabling direct access to commands when needed.

Usage typically involves describing your goal in natural language (e.g., “Fix the login bug in auth.py” or “Explore the codebase for security patterns”). The system interprets your request, activates critical skills, and coordinates a sequence of steps across agents to complete the task while preserving context for future sessions.

Prerequisites:

• Python 3.11+
• uv (package manager) https://github.com/astral-sh/uv
• Docker (for local PostgreSQL and containers)
• Claude Code CLI (if you plan to integrate Claude Code tooling)

Installation steps:

1. Clone the repository: git clone https://github.com/parcadei/Continuous-Claude-v3.git cd Continuous-Claude-v3/opc

2. Ensure prerequisites are installed and available in your environment. Verify Python and uv are ready: python --version uv --version

3. Run the setup wizard to configure the environment and MCP/server components: uv run python -m scripts.setup.wizard

4. During setup, the wizard will configure the local database (PostgreSQL via Docker by default), API keys, and clone necessary components. Follow the prompts to complete the configuration. The process will also install Claude Code integration components (32 agents, 109 skills, 30 hooks).

5. Start the server environment once setup completes. In most setups you will bring up the local services via uv and docker Compose as guided by the wizard, or run the specific startup command provided by the wizard output. If you need to uninstall later, you can run: cd Continuous-Claude-v3/opc uv run python -m scripts.setup.wizard --uninstall

6. Optional: If you’re using a remote database, adjust your environment variables or Docker Compose configuration as needed to point to your remote PostgreSQL instance, then rerun the setup or migrations as prompted by the wizard.

Notes and tips:

• The MCP configuration (mcp_config.json) is stored in the opc directory and is restored during backups. Keep a copy of your mcp_config.json and essential credentials in a secure location.
• The system uses a memory daemon to extract and surface learnings across sessions. Ensure you have persistent storage mounted for memory and history data to prevent data loss when restarting.
• If you encounter token-related issues, rely on the 5-layer code analysis and semantic indexing features to minimize token usage while preserving context.
• Environment variables commonly include API keys, database connection strings, and Claude Code integration settings. Do not expose these in public repos; use env files or secret management.
• For remote database setups, ensure Docker/PostgreSQL is accessible from the container network and that credentials are correctly configured in your environment.
• If you need to revert changes, the installer creates a timestamped archive of your ~/.claude directory and restores from backups when reinstalling.