nexus-ai-fs 0.9.9

Nexus = filesystem/context plane.

The filesystem & context plane for AI agents

Give every agent one place to read, write, search, remember, and collaborate — from a single-file script to a fleet of thousands.

Documentation · Quickstart · Examples · PyPI · Roadmap

---

Why Nexus

Every agent framework gives you tool calling. None gives you a shared filesystem. Without one, agents duplicate files, lose context between runs, step on each other's writes, and can't discover what's already been built.

Nexus fixes this. One VFS-style interface — start embedded in a single Python process, scale to a daemon-backed deployment with auth, permissions, federation, and multi-tenant isolation. No code changes.

How it works

┌─────────────────────────────────────────────────────────────────────────┐
│  BRICKS (runtime-loadable)                                              │
│  ReBAC · Auth · Agents · Delegation · Search · Memory · Governance      │
│  Workflows · Pay · MCP · Snapshots · Catalog · Identity · 25+ more      │
└─────────────────────────────────────────────────────────────────────────┘
                              ↓ protocol interface
┌─────────────────────────────────────────────────────────────────────────┐
│  KERNEL                                                                 │
│  VFS · Metastore · ObjectStore · Syscall dispatch · Pipes ·             │
│  Lock manager · Three-phase write (LSM hooks) · CAS dedup              │
└─────────────────────────────────────────────────────────────────────────┘
                              ↓ dependency injection
┌─────────────────────────────────────────────────────────────────────────┐
│  DRIVERS                                                                │
│  redb · PostgreSQL (pgvector) · S3 · GCS · Dragonfly · Zoekt · gRPC    │
└─────────────────────────────────────────────────────────────────────────┘

Kernel never changes. Drivers swap at config time. Bricks mount and unmount at runtime — like insmod/rmmod for an AI filesystem.

Get started in 30 seconds

Option A: Docker (recommended)

pip install nexus-ai-fs                       # CLI + SDK
nexus init --preset demo                       # writes nexus.yaml + nexus-stack.yml
nexus up                                       # pulls image, starts Nexus + Postgres + Dragonfly + Zoekt
eval $(nexus env)                              # load connection vars into your shell

Open http://localhost:2026. That's it.

Option B: Embedded (no Docker)

pip install nexus-ai-fs

import asyncio, nexus

async def main():
    nx = await nexus.connect(config={"data_dir": "./my-data"})

    await nx.write("/notes/meeting.md", b"# Q3 Planning\n- Ship Nexus 1.0")
    print((await nx.read("/notes/meeting.md")).decode())

    nx.close()

asyncio.run(main())

Option C: CLI

nexus write /hello.txt "hello world"
nexus cat /hello.txt
nexus ls /
nexus search query "hello" --mode hybrid
nexus versions history /hello.txt

Terminal UI

The TUI is a separate TypeScript package built on OpenTUI:

bunx nexus-tui                                         # connects to localhost:2026
bunx nexus-tui --url http://remote:2026 --api-key KEY  # connect to remote instance

File explorer, API inspector, monitoring dashboard, agent lifecycle management, and more — all from your terminal.

What you get

Capability	What it does	How agents use it
Filesystem	POSIX-style read/write/mkdir/ls with CAS dedup	Shared workspace — no more temp files
Versioning	Every write creates an immutable version	Rollback mistakes, diff changes, audit trails
Snapshots	Atomic multi-file transactions	Commit or rollback a batch of changes together
Search	Keyword + semantic + hybrid, powered by Zoekt + pgvector	Find anything by content or meaning
Memory	Persistent agent memory with consolidation + versioning	Remember across runs and sessions
Delegation	SSH-style agent-to-agent permission narrowing	Safely sub-delegate work with scoped access
ReBAC	Relationship-based access control (Google Zanzibar model)	Fine-grained per-file, per-agent permissions
MCP	Mount external MCP servers, expose Nexus as 30+ MCP tools	Bridge any tool ecosystem
Workflows	Trigger → condition → action pipelines	Automate file processing, notifications, etc.
Governance	Fraud detection, collusion rings, trust scores	Safety rails for autonomous agent fleets
Pay	Credit ledger with reserves, policies, approvals	Metered compute for multi-tenant deployments
IPC	Inbox-based inter-agent messaging via pipes	Agents talk to each other without polling
Federation	Multi-zone Raft consensus with mTLS TOFU	Span data centers without a central coordinator

All bricks and system services →

Bricks (runtime-loadable): Access Manifests · Auth (API key, OAuth, mTLS) · Catalog (schema extraction) · Context Manifests · Delegation · Discovery · Identity (DID + credentials) · IPC (pipes) · MCP · Mount · Parsers (50+ formats via MarkItDown) · Pay · Portability (import/export) · ReBAC · Sandbox (Docker) · Search · Share Links (capability URLs) · Snapshots · Task Manager · TUS Uploads (resumable) · Versioning · Workflows · Workspace

System services: Agent Registry · Agent Runtime · Event Bus · Event Log · Namespace · Scheduler (fair-share, priority tiers) · Sync · Lifecycle

Framework integrations

Every major agent framework works out of the box:

Framework	What the example shows	Link
Claude Agent SDK	ReAct agent with Nexus as tool provider	examples/claude_agent_sdk/
OpenAI Agents	Multi-tenant agents with shared memory	examples/openai_agents/
LangGraph	Permission-scoped workflows	examples/langgraph_integration/
CrewAI	Multi-agent collaboration on shared files	examples/crewai/
Google ADK	Agent Development Kit integration	examples/google_adk/
E2B	Cloud sandbox execution	examples/e2b/
CLI	40+ shell demos covering every feature	examples/cli/

Deployment options

Mode	What	Who it's for
Embedded	nexus.connect() — in-process, zero infrastructure	Scripts, notebooks, single-agent apps
Shared daemon	nexus init --preset shared && nexus up	Teams, multi-agent systems, staging
Federation	Multi-zone Raft consensus across data centers	Production fleets, edge deployments

nexus init presets

Preset	Services	Auth	Use case
local	None (embedded)	None	Single-process scripts, notebooks
shared	Nexus + Postgres + Dragonfly + Zoekt	Static API key	Team dev, multi-agent staging
demo	Same as shared	Database-backed	Demos, seed data, evaluation

# Embedded (no Docker)
nexus init                                    # writes nexus.yaml for local embedded mode

# Shared daemon
nexus init --preset shared                    # writes nexus.yaml + nexus-stack.yml
nexus up                                      # pulls image, starts stack, waits for health
eval $(nexus env)                             # load NEXUS_URL, NEXUS_API_KEY, etc.

# Demo with seed data
nexus init --preset demo && nexus up

# Add optional services
nexus init --preset shared --with nats --with mcp --with frontend

# GPU acceleration
nexus init --preset shared --accelerator cuda

# Pin to a specific version
nexus init --preset shared --image-tag 0.9.4

# Build from local source (for contributors)
nexus up --build                              # build + tag as nexus:local-{hash}
nexus up                                      # reuses local build (no pull)
nexus up --pull                               # discard local build, pull from remote

# Stack lifecycle
nexus stop                                    # pause containers (fast, no teardown)
nexus start                                   # resume paused containers (fast)
nexus down                                    # stop and remove containers
nexus logs                                    # tail logs
nexus restart                                 # down + up
nexus upgrade                                 # pull latest image for your channel

# Environment variables
nexus env                                     # print export statements for your shell
nexus env --json                              # machine-readable
nexus env --dotenv > .env                     # write .env file
nexus run python my_agent.py                  # run command with env vars injected

Docker image

Published to GHCR (multi-arch: amd64 + arm64):

ghcr.io/nexi-lab/nexus:stable          # latest release
ghcr.io/nexi-lab/nexus:edge            # latest develop
ghcr.io/nexi-lab/nexus:<version>       # pinned (e.g. 0.9.3)
ghcr.io/nexi-lab/nexus:stable-cuda     # GPU variant

Storage architecture

Four pillars, separated by access pattern — not by domain:

Pillar	Interface	Capability	Required?
Metastore	MetastoreABC	Ordered KV, CAS, prefix scan, optional Raft	Yes — sole kernel init param
ObjectStore	ObjectStoreABC	Streaming blob I/O, petabyte scale	Mounted dynamically
RecordStore	RecordStoreABC	Relational ACID, JOINs, vector search	Services only — optional
CacheStore	CacheStoreABC	Ephemeral KV, pub/sub, TTL	Optional (defaults to null)

The kernel starts with just a Metastore. Everything else is layered on without changing a line of kernel code.

Contributing

git clone https://github.com/nexi-lab/nexus.git && cd nexus
uv python install 3.14
uv sync --extra dev --extra test
uv run pre-commit install
uv run pytest tests/

For semantic search work: uv sync --extra semantic-search For Rust extensions: maturin develop --release -m rust/nexus_pyo3/Cargo.toml

See CONTRIBUTING.md for the full guide.

Troubleshooting

ModuleNotFoundError: No module named 'nexus'

Install from PyPI: pip install nexus-ai-fs. The package name on PyPI is nexus-ai-fs, not nexus.

maturin develop fails at the repo root

Point maturin at a crate manifest: maturin develop --release -m rust/nexus_pyo3/Cargo.toml

faiss-cpu resolution fails

Only install semantic search extras on platforms with compatible txtai/faiss-cpu wheels: pip install "nexus-ai-fs[semantic-search]"

License

Apache License 2.0 — see LICENSE for details.

Built by Nexi Labs.