How dkshbyte/mindbaton works
Mindbaton is a private, local-first shared memory service for conversations across AI clients. Users capture chats through the browser extension or connect agents and import existing notes; the server processes them into a searchable memory graph, which clients can query or use to create hand-off packs. The README also documents multi-account privacy, offline semantic search, optional model and provider assistance, and data export. Several server internals and UI routes were not sampled, so those subsystem links are documented or omitted rather than inferred from filenames.
Main entry points: server.py HTTP server · content.js Browser extension · mindbaton.py Agent connector · mcp_stdio.py MCP bridge
Components 19
Capture and connections
- Browser extension
extension/content.js - Agent connector
mindbaton.py - MCP bridge
mcp_stdio.py - Memory importer
import_memories.py
Memory processing
- Live sessions
live.py - Fact extraction
brain.py - Rule-gap learner
learn.py - Optional reader
deep.py - Semantic vectors
sense.py
Memory service
- HTTP server
server.py - Memory graph
server.py - Recall and answers
server.py - Optional AI providers
ai.py
AI experiences
- Handoff packs
handoff.py - Web app
index.html
Account and portability
- Private accounts
server.py - Setup and device linking
mindbaton.py
Other components
- User
- AI clients
Connections 24
- User → Browser extension: captures chats
README.md - Browser extension → HTTP server: sends captures
README.md - AI clients → Agent connector: connects agents
README.md - Agent connector → HTTP server: sends memories
mindbaton.py - AI clients → MCP bridge: uses MCP
README.md - MCP bridge → HTTP server: bridges requests
README.md - User → Memory importer: imports notes
README.md - Memory importer → HTTP server: posts memories
import_memories.py - HTTP server → Live sessions: records sessions
server.py - HTTP server → Fact extraction: processes captures
server.py - Fact extraction → Rule-gap learner: asks when needed
brain.py - HTTP server → Optional reader: requests reading
server.py - HTTP server → Semantic vectors: uses vectors
server.py - HTTP server → Memory graph: stores graph
server.py - HTTP server → Recall and answers: routes recall
server.py - Recall and answers → Memory graph: reads memories
server.py - AI clients → HTTP server: asks memory
README.md - HTTP server → Optional AI providers: requests answers
server.py - HTTP server → Handoff packs: builds handoffs
live.py - HTTP server → Web app: serves interface
README.md - User → Web app: manages memory
README.md - HTTP server → Private accounts: isolates accounts
README.md - User → Setup and device linking: links devices
README.md - Setup and device linking → HTTP server: configures connection
mindbaton.py