Long Term Memory
Entity-based knowledge graph that remembers your world.
Franklin Brain
Franklin Brain is an entity-based knowledge graph that stores information about the people, projects, and concepts you work with. Unlike session history (which is a flat log), the Brain organizes knowledge into structured entities and relationships.
Entity Types
The Brain tracks five types of entities:
- Person— people you mention or work with
- Project— codebases, repos, and initiatives
- Company— organizations and teams
- Product— tools, services, and software
- Concept— technologies, patterns, and ideas
Observations & Relations
Each entity has observations— facts with confidence scores. For example, an entity for "Alice" might have observations like "works at Acme Corp" (confidence: 0.95) and "prefers Python" (confidence: 0.7).
Entities are connected by typed relations:
founded— person founded companyworks_on— person works on projectuses— project uses concept/productpart_of— project is part of company
Automatic extraction
Exploring the Brain
Use the /brain command to explore what Franklin remembers:
# Open the brain explorer
/brain
# Search for a specific entity
/brain Alice
# Search by project name
/brain franklinStorage
The Brain is stored as JSONL files in ~/.blockrun/brain/. No external database is required. Files are human-readable and easy to back up or migrate between machines.
Brain vs. Learnings
Franklin has two memory systems that serve different purposes:
- Brain — remembers who and whatyou work with. Entities, facts, relationships. "Alice uses Next.js for the dashboard project."
- Learnings — remembers howyou work. Preferences, patterns, style. "User prefers TypeScript strict mode and 2-space indentation."
They work together