Supermemory Filesystem
https://github.com/supermemoryai/smfs- Category
- AI Agents
- Rank
- No. 815Tools index
Previous survey · No. 809 ·
- Pricing
- Open Source
- Type
- TOOL
- Builder
- supermemoryai
- GitHub
- 479 stars
- Latest release
- v0.0.5
- Date
About
Filesystem designed for AI agents. SOTA retrieval, memory profiles, sync engine, grep across PDFs, images, videos.
What it does
Supermemory Filesystem turns a remote Supermemory container into familiar working storage. Local environments mount it as a directory, while restricted runtimes receive a virtual shell backed by the same service. Agents can edit ordinary documents, use literal text search, or query by meaning when wording differs.
Why it's ranked here
The design solves a real integration problem: agents already understand files and shell commands. Local mounts also work with editors and scripts, while the virtual shell carries that model into edge and browser runtimes. The dependence on hosted ingestion and incomplete filesystem semantics keeps the verdict measured.
What's good
The interface reuses established habits instead of requiring a custom memory protocol. Flagless search inside a mount becomes semantic, while searches with flags retain literal behavior. Writes remain readable immediately through local caching, configurable scopes limit which documents enter memory processing, and unmounting drains queued uploads.
Tradeoffs
A Supermemory API key and remote service are required. Search and cross-session visibility are eventually consistent because server ingestion takes time. The virtual filesystem omits permissions, timestamps, links, device files, and large binary uploads. Multi-writer caching can briefly serve stale content, although its lifetime is configurable.
How to use it well
Use it for agents that accumulate notes, research, decisions, or working context across sessions and benefit from shell-shaped access. Mount it when editors and existing scripts need the same material. Choose the virtual shell for serverless or browser agents. Do not treat it as a complete POSIX filesystem or an offline storage layer.
Technical notes+
The Rust workspace in Cargo.toml contains crates/smfs-core and crates/smfs, uses Tokio, Reqwest, bundled SQLite, FUSE on Linux, and a pure-Rust NFS server for macOS. bash/src/create-bash.ts combines just-bash, the Supermemory SDK, SupermemoryFs, and semantic search; eager loading is enabled by default. bash/src/session-cache.ts implements a byte-limited LRU-style cache with configurable expiry. bash/src/filepath.ts requires absolute, extension-bearing paths and reserves /profile.md. bash/src/supermemory-fs.ts models directories synthetically and rejects chmod, timestamp changes, symlinks, and hard links. bash-py/supermemory_bash/_client.py supplies an async HTTP client with retries, while bash-py/supermemory_bash/_parse.py explicitly rejects command and arithmetic substitution.
Observed
- License
- MIT
- Languages
- Rust workspace, TypeScript package, and Python package implementation are present.
- Install surface
- Prebuilt CLI installer for macOS and Linux, Cargo source build, Docker image build, and TypeScript package.
- Interfaces
- Command-line filesystem mount, shell search wrapper, TypeScript library, and virtual shell tools in TypeScript and Python.
- Platform support
- macOS and Linux on arm64 and x64; virtual shell targets serverless, edge, and browser-based runtimes.
- Authentication
- Requires a Supermemory API key; the CLI can store credentials locally.
Read from README.md, Cargo.toml, bash/vitest.config.ts, bash/src/index.ts, bash/src/errors.ts, bash/src/volume.ts, bash/src/filepath.ts, bash/src/path-index.ts, bash/src/create-bash.ts, bash/src/session-cache.ts, bash/src/supermemory-fs.ts, bash/src/tool-description.ts, bash-py/supermemory_bash/_parse.py, bash-py/supermemory_bash/_shell.py, bash-py/supermemory_bash/_client.py.
What it can do
Retrieve information using state-of-the-art algorithms
Query or search terms → Relevant files and data
Create and manage memory profiles for AI agents
AI agent data and preferences → Structured memory profile
Synchronize files and data across systems
Files and directories → Synchronized filesystem state
Search text content within PDF files
Search query and PDF files → Matching text passages and locations
Search content within image files
Search query and image files → Matching visual content or metadata
Search content within video files
Search query and video files → Matching video segments or metadata
Store and organize files for AI agent access
Files and documents → Organized filesystem structure
Tags
Tech Stack
Media
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.