Toweris
Sign in

Toweris for AI agents

Use Toweris with AI help, without handing over your data

Your data stays with you

Your property info stays in your own Google account. Toweris does not copy it into a shared Toweris database.

Only the access you allow

You choose what the agent can do and how long it can do it.

You can see what changed

Toweris shows what the agent changed and when it happened.

Undo when safe

You can undo a change later if nothing else changed after it.

For assistants and crawlers, the machine-readable guide is at /llms.txt.

This page is public. You do not need to sign in to read the setup steps.

Nice things to ask when your brain is full

Calm prompts for real property work.

Home continuity lookup

Ask for the home context you usually need when something breaks, leaks, expires, or needs a handoff.

Needs: Read permission for your properties and document names/details.

The washer is acting up again. Find the model, warranty, receipt, last service notes, and any repair contact we saved.
We are leaving the house with my sister for a week. Make a simple handoff list with shutoffs, alarm notes, utility contacts, and anything weird about the home.
Before I call insurance about the roof leak, pull together roof photos, repair invoices, warranty notes, and contractor details we have saved.
The basement sink is leaking again. Show me past plumbing notes, invoices, photos, and who fixed it last time.
Which warranties would actually help if the fridge, dishwasher, or washer broke this month?
We might sell next year. Make a tidy list of upgrades and repairs with dates, costs, contractors, and attached documents.
Help me build a 10-minute home emergency cheat sheet from the records we already have.
What important home records are still missing if I want someone else to understand the house without calling me?

Read-only property lookup

Ask money questions in plain language without giving the agent permission to change anything.

Needs: Read permission for your properties.

My accountant asked for last year's property tax totals. Show me the amount for each investment property and flag anything missing.
Rent is due this week. Which properties have received the expected rent, which are short, and by how much?
Before I message anyone, show me tenants who look behind on rent and the last payment we recorded for each one.
If I raise each rental by 5%, what changes monthly and yearly, and which property moves the most?
Which property actually kept the most money last year after expenses, not just brought in the most rent?
Which property had the ugliest expense month last year, and what drove it?
How much did repairs and maintenance cost across the portfolio last year, grouped by property?
Before tax prep, which investment properties have clean T776 rent and expense totals, and which ones need review?
Which properties are losing money so far this year, and is it rent shortfall or expenses doing the damage?
Show me Pine Street month by month: rent received, expenses paid, and net cashflow.
What were my five biggest property expenses last year, and were any one-off repairs?

Document inventory without file contents

Let the agent see document names and details without giving it the files themselves.

Needs: Permission to see document names and details.

The bank asked for proof of insurance on Pine Street. Do I have the current policy document saved?
Which properties are missing insurance documents, and which ones only have older-looking files?
My tenant asked for a copy of the lease. Show me the lease-related documents saved for Pine Street.
What property documents did I add in the last 30 days, grouped by property?
Do I have a property tax bill saved for each rental before I send files to my accountant?
Show me documents that look like renewals, expirations, or notices I should deal with soon.
Which properties are missing mortgage statements or closing documents?
Show me every document tied to Pine Street grouped as lease, insurance, mortgage, taxes, repairs, and other.
Which properties still have no documents uploaded at all?
Give me a realistic upload checklist for each property, starting with the documents I am most likely to need in a hurry.

Create an entry, then review it

Let the agent add one entry, then check the result in Agent Changes and undo it if it is wrong.

Needs: Permission to add an entry, then a quick check in Agent Changes after.

Log the $1,250 Pine Street rent payment that arrived today, then point me to the exact change in Agent Changes.
Add today's $120 electric bill for Pine Street and show me where I can review or undo it.
Log the $220 plumbing repair for Pine Street with the note 'kitchen sink leak', then give me the change number.
Add the $95 HOA payment for Pine Street dated today, then take me to the review page.
Log the $2,000 insurance payment for Pine Street dated January 15, 2026, then show me how to check it.
Add the $48 water bill for Pine Street dated March 1, 2026, then point me to Agent Changes.
Log yesterday's $140 move-out cleaning expense for Pine Street, then show me where I can undo it if needed.
Add a $300 maintenance expense for Pine Street dated today with the note 'patched stair railing', then show me the change.
Log the $60 trash bill for Pine Street dated today and point me to the change number.
Add today's $75 lawn care expense for Pine Street, then show me the review page before I move on.

What this setup is for

Use this if you want AI help in Toweris.

Good for

  • Connecting ChatGPT, Codex, or OpenClaw with one hosted sign-in flow.
  • Letting the AI add or edit things you allow.
  • Checking what the AI changed after it runs.
  • Keeping setup to a few simple steps.

Not for

  • Giving the AI full access forever.
  • Sending your data into a shared Toweris database.
  • Hidden changes you cannot review.
  • Using the wrong Google account or workspace.

What you need

Keep it simple.

  • Open your Toweris workspace in the browser once first.
  • Use an assistant that supports hosted MCP OAuth sign-in, like ChatGPT, Codex, or OpenClaw.
  • Use the same Google account everywhere.
  • Finish the browser sign-in on the same machine where the assistant is running.

Pages in this flow

Start here.

  • /agents: public setup guide.
  • /llms.txt: public plain-text guide for AI tools.
  • /agent-changes: sign-in required page to review changes and undo safe ones.

First-time setup

Do this once.

1) Open your workspace once in the browser

This tells Toweris which account you want to use.

Sign in with Google to open or setup your own private workspace.

2) Connect your assistant

Use the same hosted Toweris MCP URL in any client that supports MCP OAuth.

  1. Add the Toweris MCP in your assistant.
  2. Use this web address:
  3. https://staging.toweris.com/api/mcp
  4. Sign in with Google when prompted.
  5. Finish that browser flow on the same machine where your assistant is running.
  6. Open the same Toweris workspace you already used.

This is the same hosted OAuth flow for ChatGPT, Codex, OpenClaw, and any other MCP client that supports browser sign-in and the local callback handoff.

3) Approve limited access

Only approve what you want the AI to do.

  • Only approve the actions you want.
  • Keep access short.
  • Check Agent Changes after changes.
  • Undo mistakes from the same page when it is safe.

4) Finish on the same machine as your assistant

This is the part that feels weird if a client handles callbacks badly.

  • Finish Google sign-in in a browser on the same machine where your assistant is running.
  • Do not move the last browser step to your phone or another computer.
  • If the last page lands on 127.0.0.1 or localhost, that is usually the assistant's local callback step.
  • If that page does not load, Toweris may already be done and the assistant may have stopped listening too early.
  • Start a fresh login attempt if that happens.

How you can see agent changes

You should always be able to see what changed.

  1. Every AI change is saved with its name and time.
  2. Each change gets its own change number.
  3. Open Agent Changes to review changes and undo safe ones.
  4. Toweris may block undo if something changed later.

Restrictions and guardrails

These are the safety rules.

  • Toweris does not copy your data into a shared Toweris database.
  • Agents get no access unless you allow it.
  • For now, agents can only see document names and details, not the file contents.
  • If the wrong account or workspace is open, Toweris blocks the request.
  • If something changed later, Toweris may block undo to keep your data safe.

FAQ

Does Toweris upload my data so an agent can use it?

No. Your data stays in your own Google account.

Does the no-install hosted setup work for every assistant?

No. It works for assistants that support hosted MCP OAuth sign-in and can finish the browser callback on the same machine.

Can an agent see everything once I connect it?

No. You choose what the agent can do. If you did not allow it, Toweris blocks it.

How do I see what an agent changed?

Open Agent Changes. It shows what the agent changed.

Can I undo an agent mistake?

Yes, if it is still safe to undo.

What if Google sign-in works but the last localhost page does not load?

That usually means Toweris finished correctly, but your assistant stopped listening for the local browser callback too early. Start the login again and finish it on the same machine and browser session as the assistant.

If something breaks

  • DELEGATION_TOKEN_REQUIRED: the app did not send the sign-in proof Toweris needs.
  • DELEGATION_AGENT_ID_REQUIRED: the app did not say which agent is calling.
  • DELEGATION_TOKEN_REPLAYED: that token was already used. Try again.
  • DELEGATION_TOKEN_INVALID: that sign-in proof is not valid.
  • DELEGATION_TOKEN_EXPIRED: that sign-in proof expired. Try again.
  • DELEGATION_REVOKED: that access was turned off.
  • DELEGATION_TOOL_NOT_ALLOWED: the AI was not allowed to do that.
  • DELEGATION_SCOPE_MISSING: the AI does not have the needed permission.
  • DELEGATION_RESOURCE_FILTER_DENIED: the AI was not allowed to touch that property.
  • BYOS session is not initialized: open your workspace again, then try again.
  • FOREIGN KEY constraint failed: Toweris could not find that property in this workspace.
  • Context mismatch: the wrong account or workspace is connected.
  • Last browser step lands on 127.0.0.1 or localhost and does not load: Toweris may have already finished, but the assistant likely stopped listening for the local callback too early. Retry from a fresh login attempt on the same machine.