ai-continuity-plugin

The ai-continuity-plugin offers a durable solution for managing project states across AI sessions—keeping track of critical decisions, constraints, and rejected options in a version-controlled manner via Git. This is distinct from traditional memory tools, focusing instead on maintaining the present state of a project and what should remain unchanged.

Key Features

• Markdown-Based State Management: The project state is stored as plain markdown files, which are tracked through Git. This allows for an easy and human-readable way to review changes, rollback decisions, and conduct audits. For example, project history can be tracked using the git log command.

• Flexible Storage Modes: Two distinct storage modes are available:

  • Repo Mode: For environments with a project working directory, such as Claude Code.
  • Central Mode: For tools without a project working directory, allowing project management by name.

• Deterministic Context Resumption: The context resumption for AI sessions is derived from markdown files, devoid of LLM influence. This means users can see confirmed constraints, decisions, and open questions as they continue their projects.

• Transparency and Confidence Tracking: Each decision made carries confidence levels and provenance, ensuring that users understand the reliability of information being presented. This mitigates reliance on unverified inferences, establishing a trust-based environment for AI interactions.

Usage Example

The command-line interface offers intuitive commands to manage project states:

continuity init "Build Snip, an internal URL shortener"
continuity record-decision "PostgreSQL is the source of truth" --body "ops knows it; volume is modest"
continuity reject "DynamoDB as primary store" --reason "KV overhead unjustified at our volume"
continuity save "7-char base62" # lock a constraint
continuity resume # retrieve the current project state
continuity why "301" # see the rationale behind a decision

Compatibility and Installations

The plugin can be installed as part of the Claude Code ecosystem, functioning seamlessly across various platforms including terminal, desktop, and web applications. The setup process is straightforward, utilizing specific commands to integrate the plugin into the working environment, ensuring efficient capture and management of project states.

Conclusion

The ai-continuity-plugin sets a new standard in AI project management, moving beyond basic memory capabilities to offer a robust, version-controlled environment where every aspect of project continuity is documented and accessible. This ensures that projects can progress without the risk of losing critical insights and decisions, fostering a reliable collaborative workspace.