Give your AI agents secure, real-time access to the data in your LucidLink filespace. This repository shows
how to connect popular Python agentic frameworks to LucidLink - either through the
Python SDK (wire filespace list / read / write in as agent
tools) or the MCP server (drop in 47 filespace tools).
Each framework integration gets its own folder under frameworks/, with tiny,
self-contained files - typically a via_sdk.py (hand-wires the SDK as tools) and a via_mcp.py
(bridges to the MCP server), plus a quickstart.py where it helps.
→ Browse, set up, and run them from the catalog: frameworks/README.md.
Looking for other LucidLink AI integrations? LucidLink AI is the place - it indexes all of them.
Let your agent mint collaborator service accounts - machine teammates that see nothing until granted - and hand a worker agent exactly one folder of your filespace, optionally on a self-destructing single-use credential. All in natural language through the MCP.
→ Walk through it: scoped-agents/README.md.
Three LangGraph agents share memory as files on a filespace: one folder per namespace, one file
per memory, an exclusive lock on every update, and an audit trail that attributes each write to
the agent's own service account. A BaseStore adapter over the Python SDK, a two-take collision
demo, and three agents with three identities and three scopes.
→ Run it: fleet-memory/README.md.
Wire terminal AI agents - OpenClaw, Hermes, and NVIDIA NemoClaw sandboxes - at your filespace through the MCP, then read and write files in natural language.
→ Set them up: claws/README.md.