Skip to content
Nivlo
Agentic infrastructure · early access

Your agents can think.
Now let them act.

Today's AI assistants are brilliant and useless at once — they can reason about your work, but they can't touch a single file. Nivlo is the secure layer that connects any agent — Codex, Claude Code, your own — to your real content, to find, organize and deliver it on your behalf. You stay in control. Your data stays in the EU.

See how it works

Open MCP standard · Scoped permissions · Every action mirrored & reversible

MCP-native
One standard endpoint
500K+
Files agents can reach
100% EU
Data on dedicated servers
Glass-box
Every action, visible

Works with OpenAI Codex, Claude Code, Hermes and any MCP client. Built on the Nivlo platform, already trusted by companies and agencies.

Why now

Every company is about
to run agents.

Almost none of them can safely act on the company's own files. That gap is the whole opportunity.

Agents got smart. Access stayed dumb.
Models can reason about your work — but they can't open, move, or deliver a single asset without a human doing it for them.
A chatbot that can't touch your files is a demo.
Value shows up when the agent does the task — not when it describes what you should do next.
The bottleneck isn't intelligence. It's permission.
Acting on real data demands scope, audit and trust. That's infrastructure — not a longer prompt.
What it means for you

Your work, done
by your agents.

Not someday. The moment you connect one. Agentic infrastructure isn't a feature — it's the difference between an assistant that talks and one that delivers.

Hours → seconds
“Find the approved 2026 logo, tag the new shots, send the press kit.”

The work you'd assign to a junior, done before you finish the sentence — across thousands of files, without you opening a single folder.

Your files become callable
Every asset turns into a tool your agents can use.

Search, read, move, tag, share — your content stops being a passive archive and becomes something an agent can actually operate.

You stay in command
Nothing happens behind your back.

Every action is visible live, sensitive steps wait for your approval, everything is reversible — and nothing ever leaves the EU.

Bring your own agent
Codex, Claude Code, Hermes — or the one you built.

No custom integration, no lock-in. If it speaks MCP, it works with your files today.

How it works · MCP

Connect an agent.
It gets hands.

An external agent only becomes useful when it can touch your real files. Nivlo gives it exactly that — through a single MCP endpoint, with the permissions you choose.

01
Connect

Give your agent Nivlo’s MCP endpoint and a token scoped exactly how you want — read-only or read-write, on the sections you allow.

02
It gets tools

The agent instantly receives Nivlo’s tools: search, read, move, tag, share files. No custom integration, no glue code.

03
It acts on your behalf

Every call is permissioned, mirrored live in the Nivlo UI, and reversible. Sensitive steps wait for your approval.

Compatibility

Works with any agent
that speaks MCP

MCP is an open standard. If your agent supports it, it connects to Nivlo — no separate integration built for each one.

OpenAI Codex
CLI & API agent
via MCP
Claude Code
Terminal agent
via MCP
Hermes
Personal assistant
via MCP
Cursor
Agentic IDE
via MCP
Claude Desktop
MCP client
via MCP
ChatGPT
Connectors / MCP
via MCP
Your own agent
Any language, any SDK
via MCP
+ any MCP client
Open standard
via MCP
Tools

What an agent can do
with your files

The same operations you'd perform in Nivlo — now available to your agent as MCP tools, under your permissions.

In active development · rolling out gradually
01 — Access & search

Find the right file,
without looking for it

The agent navigates your structure, searches by name, type, tags or content, and reads the file or its preview. Read-only tools — safe from the very first connection.

  • List folders and sections
  • Search by name, type, tag, date
  • Read content and metadata
  • Preview 50+ file types
files.search · files.read · files.list
02 — Organize & act

Put things in order,
not just read them

The agent moves, renames, restructures folders, tags and versions. Write operations — all reversible and logged. Nothing happens off the record.

  • Move, rename, restructure
  • Tagging and metadata at scale
  • Automatic versioning on change
  • Every write — reversible
files.move · files.tag · files.version
03 — Distribute

Deliver,
not just arrange

The agent creates share links with passwords and expiry, sends large files via Nivlo Send, and prepares social-media formats — on your command.

  • Share links with password & expiry
  • Nivlo Send — large files, no account
  • Social-media formatting
  • Differentiated access per recipient
share.create · send · format
04 — Permissions & glass-box

You own every
action the agent takes

Clearly scoped tokens apply to every agent. Each call is mirrored live in the interface, logged and reversible. Sensitive steps wait for your explicit approval.

  • Scoped tokens per agent and section
  • Live mirror + full audit trail
  • Mandatory approval on sensitive steps
  • Instant access revocation
[ Per-agent permissions & audit panel ]
For technical teams

An MCP server
you can build on

One endpoint, typed tools, granularly scoped tokens, multi-tenant isolation and EU hosting.

MCP endpoint

A single URL: api.nivlo.com/mcp. Any compatible client connects — no custom integration.

Scoped tokens

Granular permissions per token and agent: read-only or read-write, limited to sections. Revocable anytime.

Typed tools

Clear schema: files.search, files.move, share.create. Same tools over the API.

Multi-tenant isolation

An agent only sees its own tenant. One organization's data and actions never touch another's.

Real-time mirror

A live event channel (websockets). Every agent call is streamed into the UI and logged.

EU · GDPR · self-host

Dedicated EU servers. On Enterprise: self-hosted, SLA and a dedicated account manager.

Why Nivlo is the layer to bet on

Open MCP standard
Your agents, not just ours. No lock-in, no per-agent integration to maintain.
On your behalf, in control
Scoped tokens, approval on sensitive steps, every action reversible.
Glass-box
See every agent call live in the UI. Full audit, zero black-box.
EU & tenant-isolated
Data on European servers, isolated per organization. GDPR-native.

Give your agents hands.

We're onboarding design partners — teams already running agents who want them working on real files. Tell us your stack, get in the first wave.

or explore the Nivlo platform