Continuous Claude
An environment for Claude Code that keeps context across sessions via hooks, ledgers and handoffs, orchestrates agents and saves tokens
High risk
We rate an entry high when the tool writes to external systems, handles money, production databases or secrets, or runs arbitrary commands. The CLI installs it only with your consent.
Why this level
- Installs hooks that run commands on agent actions and changes the global config in ~/.claude
- Deploys a PostgreSQL database and a daemon, and handles API keys
Install
Manual install
git clone https://github.com/parcadei/Continuous-Claude-v3.git
cd Continuous-Claude-v3/opc
uv run python -m scripts.setup.wizardThe twelve-step wizard installs the integration, a PostgreSQL database in Docker and the tools. Run uv commands from the opc directory.
This is third-party code. Review the repository files before installing.
What it does
Continuous Claude tackles context loss on compaction in Claude Code: instead of squeezing the history, the system extracts a session's learnings and starts fresh with context already assembled. Input hooks suggest which skills and agents fit a request, maintain state ledgers and write cross-session handoffs in YAML. It ships a set of skills, specialized agents and hooks, plus a layered code analysis that hands the model structure instead of reading whole files. Memory accumulates learnings through a daemon and surfaces them at startup. It installs via a setup wizard that deploys the integration into the .claude directory, a PostgreSQL database in Docker and helper tools.
Who it is for. For developers who work heavily in Claude Code and want to keep context and learnings across sessions.
Good fit when
- Context in Claude Code runs out and key decisions are lost on compaction
- You need to resume work from where the previous session stopped
- You need to orchestrate several agents with isolated context windows
- You want to spend fewer tokens reading large files
Not a fit when
- You cannot run Docker and PostgreSQL for the local store
- You need a light single skill without changing the global Claude Code config
Example request
I am done for today, save the state so I can resume from here tomorrowLimitations
It works only with Claude Code. Installing needs Python 3.11+, the uv package manager, Docker for PostgreSQL and the Claude Code CLI itself, and uv commands run from the opc directory. The wizard changes the global config in ~/.claude; it makes a backup before installing and has an uninstall mode for rollback. Perplexity and NIA are optional and require their own keys.
How to disable. Run the wizard with the uninstall flag: from the Continuous-Claude-v3/opc directory run uv run python -m scripts.setup.wizard --uninstall. It archives the current setup and restores your previous Claude Code config from the backup.
Security check
- Installs hooks that run commands on agent actions and changes the global config in ~/.claude
- Deploys a PostgreSQL database and a daemon, and handles API keys
README in short
The README describes Continuous Claude as a learning, multi-agent environment on top of Claude Code that keeps context across sessions and saves tokens. Its core is skills, agents and hooks, layered code analysis, a memory system, continuity ledgers and a coordination layer. An input hook suggests the right skills and agents by priority level, and natural language replaces memorizing slash commands. Installation runs through a twelve-step wizard: config backup, environment check, database, migrations, integration of agents, skills and hooks, and extra tools. It has update and uninstall modes, remote PostgreSQL setup and environment variables. MIT license.
FAQ
Is a database required?
Yes, memory and the semantic index are stored in PostgreSQL. By default the database runs locally in Docker, but you can connect a remote one with the pgvector extension.
Are paid services required?
No paid services are required. Perplexity and NIA are optional, improve search quality and need their own keys.
Related
Open-source personal AI assistant on your own machine: answers in Telegram, Slack, Discord and WhatsApp, extended with skills and plugins
A self-improving agent from Nous Research with a TUI, messaging gateway, cron jobs and skills it writes itself
An open source coding agent for the terminal and desktop with build and plan modes
Open prompt library with a Claude Code plugin, MCP server and CLI: search, fetch and improve prompts and skills from an agent