Pensieve
tore your decisions and principles. Claude reads them to make better choices.
Pensieve is an AI agent skill that builds a self-growing, four-layer knowledge base for project context and decision history. It captures engineering principles, architectural decisions, exploration results, and workflows automatically as you develop, replacing static CLAUDE.md files with a semantic knowledge graph. The skill routes relevant context on-demand to minimize token usage while helping agents make consistent, project-aligned decisions. Compatible with Claude Code, Cursor, and other AI clients that support skills, it turns daily development into accumulated institutional memory.
Key Features
Use Cases
- 01Validate AI-generated implementation plans against stored architectural conventions before execution
- 02Locate module entry points and boundaries by reusing previous exploration results without global search
- 03Accumulate code review lessons so standards improve over time instead of repeating mistakes
- 04Analyze refactoring impact by following semantic links to discover affected workflows and dependencies
- 05Commit code using project-specific conventions and decisions without re-asking style questions
- 06Maintain consistent engineering philosophy across team members and long project timelines
Related Skills
View moresuperpowers
An agentic skills framework & software development methodology that works.
skills
Skills for Real Engineers. Straight from my .agents directory.
skills
Public repository for Agent Skills
ponytail
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
Pensieve — FAQ
What is Pensieve?+
Pensieve is an agent skill that creates a structured, self-growing knowledge base for AI assistants, storing project conventions, architectural decisions, exploration results, and workflows. It replaces static CLAUDE.md files with a semantic graph that agents query on-demand for minimal context usage.
How do I install Pensieve in Claude Code?+
Clone the skill globally to ~/.claude/skills/pensieve, run the hook installer, then initialize per-project data using the provided init script. Each project maintains independent knowledge while sharing the same system code.
Which AI clients support Pensieve?+
Pensieve works with Claude Code (with hooks for auto-sync), Cursor, and any AI agent client that supports skills. Installation paths differ by client, but the four-layer knowledge model and tools remain the same.
Does Pensieve require API keys or external services?+
No, Pensieve runs entirely locally using git, bash, and Python 3.8+. It stores all knowledge in your project's .pensieve directory without calling external APIs.
Is Pensieve free to use?+
Yes, Pensieve is open-source under the MIT license and completely free. You can use, modify, and distribute it without restrictions.
How does Pensieve accumulate knowledge automatically?+
The self-improve tool extracts insights from your conversations and code diffs, writing them to the appropriate knowledge layer. Claude Code hooks trigger graph sync after edits; other clients can manually run self-improve.
How do I install Pensieve?+
Open the source repository on GitHub and follow its README. Pensieve is a skill — MCP Agents Market links you directly to the official repo.
Is Pensieve free?+
Pensieve is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.