Netridium — Client Manager Guide (for Service Providers)

For MSPs, accounting firms, and consultancies who use the Client Manager to build, deliver, and govern AI tools across many client companies from one place.

The model in one line: your firm builds a capability once and delivers it to every client you manage — each client a fully isolated workspace, all run from a single console. Your staff have full admin access to every client you manage, so you can set up and operate a client's AI assets on their behalf.

This guide is for the provider role (the "Client Manager"). If you're setting up or running tools as a single workspace, see CLIENT_ADMIN.md (builders) and USER_GUIDE.md (end users).

1. How the tiers fit together

Netridium has three kinds of workspace:

*Key idea — you work inside each client's workspace. Your firm's console (Clients, Staff, Audit, Settings) is for managing* clients. The actual AI assets — AI Tools, connectors, roles, users — live in each client workspace, and you build and manage them by opening that client (you have full admin access there). Your own provider workspace has no AI Tools of its own.

2. Your console

When you sign in to your firm's workspace, the header reads "… Client Manager" and the sidebar has four items:

3. Onboarding a new client

From Clients → New client, fill in:

On save, Netridium:

1. Creates an isolated client workspace linked to your firm.
2. Creates two built-in roles in it — Admin and User.
3. Emails the admin invite so the client can set a password and join.

You don't have to wait for the client to accept — your staff already have admin access to the new workspace and can start setting it up immediately.

4. Opening and working inside a client

On the Clients list, each client row shows its Company, Mode, and Status, with actions:

• Open — enters the client's workspace (in a new tab). Because your staff are

  admins of every client you manage, you land with full access and can do anything a client admin can: build AI Tools, connect systems, manage roles and users, run tools, and review history.

• Edit — opens the client's profile (name, product, mode, address) plus the

  Require BYOK control (see §7).

• Suspend / Activate — pause or resume a client without deleting anything.

Inside an opened client workspace you'll see the full client sidebar: My AI Tools, AI Tools (build), History, Connectors, Roles, Users, Audit Log, and Settings.

5. Connecting the client's systems

Every client connects its own accounts; credentials are encrypted and never shared across workspaces. Open the client → Connectors and add one per system:

• Available connectors: NetSuite, Salesforce, Maxio, QuickBooks Online, Xero,

  Google Drive.

• OAuth connectors (Salesforce, QuickBooks Online, Xero, Google Drive) — click

  Connect and authorize in the provider's login screen; no copying API keys.

• Endpoint connectors (Maxio, NetSuite) — provide the connection URL and token.
• Read-only by default. Grant write access explicitly, per connector, only when

  a tool genuinely needs to change data — a tool can never modify data through a connector that isn't write-enabled.

Provider-by-provider steps are in CONNECTOR_SETUP_FOR_ADMINS.md. Some connectors need a one-time platform OAuth-app setup by the System Manager — see §8.

6. Build once, deliver to many

This is the heart of the Client Manager model. You build a tool against deep domain knowledge once, then deliver it to every client that should have it.

Build a tool (inside a client)

Open a client → AI Tools → New, give it a title, pick the connector(s) it may use, and choose a type:

• Prompt (natural language) — describe the outcome; the AI works out the steps.

  Read-only.

• LSPL (LogicScript Prompt Language) — write the steps as a structured script

  the AI executes in order. Repeatable and explicit. Read-only.

• Script (Python) — deterministic logic in a secure sandbox; the only type

  that can write data, and available only where Python is enabled (see §7).

Use Generate to have Netridium draft a runnable script from a plain-language description, and add an input form so end users just pick values and click Run. Full detail: LOGICSCRIPT_TYPES.md and PROMPT_INPUTS.md.

Tip — keep a template client. Since tools live inside client workspaces, many firms keep one Sandbox/Demo client as a build-and-stage library, then export from it and import into each production client.

Deliver it with export / import

Inside a client, the AI Tools page has Export and Import:

1. Export downloads a portable file of that workspace's AI Tools —

   each tool's title, type, text/script, input form, options, and its connector references (by name + type, not internal IDs, so they re-link in any workspace).

2. Import into another client uploads that file and recreates the tools there.
   • Connectors are matched to that client's own connectors by name + type;

     any that don't exist yet are left unlinked and reported (set them up, or rename to match), so the import never hard-fails.

   • A tool whose name already exists is skipped (safe to re-import).
   • Python Scripts are skipped for any client that doesn't have Python enabled.

This is how a single, well-built library propagates to many clients while each client keeps its own connections and data.

7. Governance you control per client

Most AI-asset settings are made inside the client workspace (where you act as its admin). A few are gated higher up. Here's who can set what:

Two provider-specific controls worth calling out:

• Require BYOK. On a client's Edit screen you can require that client to run

  on its own Anthropic key — their runs then never fall back to the shared platform key. Useful for cost isolation and data-flow control. The client enters their key in their own Settings.

• Enable Python. Python is the only type that can write data, so it's an

  opt-in capability granted by the platform operator, not by you. You'll see the "Enable Python" setting on a client's Settings page, but it's read-only for you — ask the System Manager (Netridium) to enable it for a specific client. Until then, Prompt and LSPL (both read-only) are available everywhere.

8. Publishing tools to client users

End users should see only the tools meant for them. Inside a client:

1. Roles — create roles (e.g. "Finance", "Sales") and, per role, choose which

   AI Tools its members can see (View) and run (Run). Run implies View. The built-in Admin role can run everything.

2. Users — invite client users by email and assign each a role. New members can

   be required to be approved before they gain access.

A user then sees only their authorized tools under My AI Tools — never the prompts or code behind them. Change or revoke a role and it updates for everyone in it at once.

9. Managing your firm's staff

Open Staff to manage your own team:

• Invite a teammate by email — they get an email to set a password and join.
• Everyone on Staff is an admin of the firm and of all its client companies, so

  add staff deliberately. Pending invitations can be resent or revoked.

There are no per-tool roles for staff — staff access is firm-wide by design (that's what makes "open any client and manage it" work).

10. Monitoring, cost & audit

• Per-client history — open a client → History to see every run, its

  inputs, result, and cost. Re-open or re-run any past report.

• Usage & cost — each run records token usage and an estimated cost, so spend is

  transparent per client and clean enough to attribute or bill on. Set Require BYOK (§7) to push a client's AI spend onto their own key.

• Audit log — both your firm's console and each client workspace keep an audit

  trail of who did what, against which data, and when.

11. Client lifecycle

• Suspend / Activate — from the Clients list, pause a client (e.g. non-payment)

  and resume later without losing anything.

• Edit — update a client's name, product, mode, profile, and Require-BYOK.
• Delete — from a client's Edit screen, permanently remove the client and all of

  its data. This cannot be undone — suspend instead if you only need to pause access.

12. What to request from the System Manager (Netridium)

A few things are operator-controlled. Ask the System Manager when you need them:

Related guides

• CLIENT_ADMIN.md — building and managing tools within a single

  workspace (everything you do inside a client).

• USER_GUIDE.md — for the end users who run the tools.
• LOGICSCRIPT_TYPES.md — Prompt vs LSPL vs Script, in depth.
• PROMPT_INPUTS.md — building self-service input forms.
• CONNECTOR_SETUP_FOR_ADMINS.md — connecting each

  system, step by step.

• FEATURES.md — the full feature overview.