Your models never leave your machine.

Last updated: July 28, 2026

This page summarizes BIM Monkey's security posture for firms, IT reviewers, and third-party risk reviews. It explains what stays on your machine, what travels to our servers, where your data lives, and how we protect it. For security questionnaires or detailed documentation, contact [email protected].

How to read this page: some protections depend on your plan — where your data is stored (Data Residency, on Pro and Enterprise) and where AI processing runs (Private AI, on Enterprise). Statements that differ by plan are labeled. Anything unlabeled applies to every plan, always: no training on your data, no cross-firm sharing, and your Revit model file never leaves your machine.

At a glance

Cloud infrastructure	Railway (AWS, United States) for API and database; Netlify for the web app
Primary data types	Drawing page images, generation history, firm standards, and project metadata. Native Revit models are never uploaded — they stay on your machine.
Data residency	BIM Monkey-hosted (US) by default. On the Pro and Enterprise plans, your firm's data can live in a dedicated database and storage bucket you control, in the region you choose — including your own cloud tenancy. Self-service from Settings.
AI processing	Anthropic's Claude API (SOC 2 Type II) by default. On the Enterprise plan, Private AI runs Claude inside your own AWS account (Amazon Bedrock) or Google Cloud project (Vertex AI) — drawing content never touches a shared AI API, and inference bills to your cloud invoice.
Firm isolation	Strict per-firm isolation; no data is shared or commingled across firms. Optional physical separation via a dedicated per-firm database and storage bucket.
AI training	BIM Monkey does not use your data to train AI models, and Anthropic does not train on API data by default
Encryption	TLS 1.3 in transit; AES-256 at rest; customer database and storage credentials encrypted with AES-256-GCM
Human in the loop	No model changes without your explicit instruction and approval
Security contact	[email protected]

Your Revit models never leave your computer

The BIM Monkey plugin communicates with Banana Chat through a named pipe on your local machine. When you issue a command, what's sent for AI processing is your instruction text, structured model data (element IDs, parameter names, sheet numbers, model summaries), and any view captures you explicitly ask it to analyze. Your actual Revit model file — geometry, families, materials, the .rvt itself — never leaves your computer, on any plan.

We do not train on your data

BIM Monkey does not use your projects, Revit models, drawing content, or Banana Chat conversations to train AI models. Your firm's data is yours. See Terms of Service §3 for the full statement on data use.

Encryption in transit and at rest

All communication between the Revit plugin and BIM Monkey's servers is encrypted using TLS 1.3. On BIM Monkey-hosted storage (the default), data at rest — including drawing uploads, model snapshots, and generation history — is stored on Railway's infrastructure using AES-256 encryption. Firms using customer-hosted data residency store this data in their own infrastructure, where at-rest encryption is provided by their chosen platform (AWS RDS/S3 and Supabase both encrypt at rest by default); the credentials you give BIM Monkey to reach it are encrypted with AES-256-GCM.

Every change requires your approval

Banana Chat does not modify your Revit model without your explicit instruction. You read what it plans to do, and you decide whether it runs. There are no background automations or model changes without a human in the loop.

AI Governance — your admins decide what the AI can do

Firm administrators control what the AI may do across the whole firm. Script execution can be restricted or disabled firm-wide from Settings → AI Governance — enforced at the platform level, not by per-user discipline. A policy change is set once by your admin and applies to every seat.

Platform automation — curated scripts and skills — is opt-in, version-pinned, and recallable: nothing your firm hasn't explicitly accepted can run, you can read the full code or instructions of any item before accepting it, and what your firm approved is exactly what runs — any update requires re-acceptance. If BIM Monkey recalls an item, further use is blocked and your firm is notified; a recall never modifies or removes anything on your side.

Every script execution is logged: what ran, who ran it, when, and its exact content signature — including attempts that were blocked by policy. The full activity log is visible to your firm and exportable as CSV for security review.

Firm data isolation

Your drawing library, generation history, and firm-specific standards are strictly isolated to your account. BIM Monkey does not share data between firms. No firm's corrections, uploads, or outputs are used to improve another firm's results.

Data residency — your data, your database, your region

By default, firm data is hosted on BIM Monkey's infrastructure in the United States. Firms with data-locality or isolation requirements can go further: your firm's data can live in a dedicated Postgres database that you control — in the region you choose, or entirely inside your own cloud tenancy. This is physical separation, not just logical: your scripts, skills, generations, corrections, and project data move to your database, and BIM Monkey's shared infrastructure no longer stores them. Data residency is included on the Pro and Enterprise plans.

• Bring your own database — your IT team provisions Postgres in your own cloud account (e.g. AWS RDS in your required region); BIM Monkey connects with credentials you control and can revoke. Configured self-service from Settings → Data Residency. You verify the database and storage connections with one-click tests before any data moves, and the migration verifies row counts before removing anything from BIM Monkey's infrastructure.
• No IT team? No problem — a managed Postgres provider such as Supabase works the same way: create a project in your chosen region (about ten minutes, with a step-by-step guide built into Settings → Data Residency), and your firm owns the account, the credentials, and the database. BIM Monkey never holds the keys to your data — you do.
• What stays with BIM Monkey — your account, login, billing, and anonymous usage telemetry. Short-lived caches (under 5 minutes) may briefly hold data in transit.
• Drawing images too — drawing page images can move to your own S3-compatible storage bucket (AWS S3, or the storage in your Supabase project), so both your database and your files live in your infrastructure.
• What data residency does and doesn't cover — data residency controls storage, not AI processing. On the Standard and Pro plans, AI generation sends drawing content to Anthropic's API for processing wherever your data is stored. For firms that need AI processing inside their own tenancy too, the Enterprise plan adds Private AI: Claude runs in your own cloud via Amazon Bedrock (AWS) or Vertex AI (Google Cloud), configured self-service from Settings → Private AI. Inference bills to your own cloud invoice, and a built-in test verifies your cloud setup before any routing changes. Private AI is independent of storage residency — a firm can adopt either or both.
• Database and storage credentials you provide are encrypted at rest with AES-256-GCM and are never displayed again after saving — not even to your own account.

What data we collect and why

We collect the minimum necessary to run the service:

• Plugin usage metadata — which tools were called, call duration, success/failure. Used to monitor reliability. Message text is never collected.
• Workflow signals — which views you activated, which element categories were modified, session timing. Used to build firm-specific working standards for your account only.
• Drawing uploads — PDF page images sent for AI analysis: to Anthropic's API on Standard and Pro, or to your own cloud tenancy on Enterprise with Private AI. Not shared with other firms. Not used to train models.

Full details are in our Privacy Policy, including a complete list of third-party processors.

Anthropic API

On the Standard and Pro plans, Banana Chat uses Anthropic's Claude API to process your instructions. Under Anthropic's standard API terms, inputs and outputs may be retained for up to 30 days for trust & safety review, after which they are deleted. Anthropic does not use API data to train models by default. Anthropic is SOC 2 Type II certified. Enterprise Zero Data Retention agreements are available directly from Anthropic for firms that require them.

Private AI — Claude in your own cloud (Enterprise)

On the Enterprise plan, AI inference doesn't go to a shared API at all. Claude runs inside your firm's own AWS account via Amazon Bedrock, or your own Google Cloud project via Vertex AI — under credentials you create, scope, and can revoke. Drawing content sent for AI processing stays inside your cloud tenancy, and inference costs bill directly to your own cloud invoice with no markup.

• No shared-API exposure — some vendors promise to limit what they send to shared AI APIs. With Private AI enabled, BIM Monkey doesn't send drawing content to a shared API at all.
• Model providers never receive your data — under AWS's and Google Cloud's service terms, Bedrock and Vertex AI inputs and outputs are not shared with the model provider (Anthropic) and are not used to train or improve any model. Your prompts are processed in your tenancy and not stored after the request completes; invocation logging is off by default and controlled entirely by you.
• Hard-fail routing — once enabled, all of your team's inference routes to your cloud. If your configuration breaks, requests fail with a clear error; they never silently fall back to a shared API.
• Covers every surface — uploads, generation, web Banana Chat, redlines, and Revit Banana Chat (plugin v0.4.20260707+). Enterprise workstations don't need an Anthropic API key.
• Self-service verification — built-in setup guides for AWS and Google Cloud, one-click inference tests, and per-model entitlement checks before anything activates.
• Independent of Data Residency — you can run Private AI without hosting your own database; all you need is a cloud account.
• Extends to AI agents — Agent Access calls authenticate per-firm, honor your data residency, and any inference they trigger on our platform runs in your cloud. Note: the AI assistant you choose to connect (Claude, Cursor, Gemini) processes the data it retrieves under your agreement with that vendor — connect assistants your firm has approved.

Why firms choose Private AI

An AI still processes your drawings — that's the product. What changes is which infrastructure the content travels through, and who can retain it. Three reasons firms make the switch, in the order they usually raise them:

• Retention — a breach or subpoena at a shared AI vendor can't expose what was never stored there. Under Private AI, prompts are not retained after the request completes, and any logging is a switch your firm controls in your own cloud account.
• Training — no training on your data is already contractually excluded on every plan. But "the model provider never receives your content at all" ends the conversation entirely.
• Procurement speed — data processed under your firm's existing AWS or Google Cloud agreement often skips the new-vendor security review that a shared-API tool triggers. Private AI isn't just safer on paper — it's frequently faster to approve.

Who can see your drawing content

On every plan: no one trains on your drawings, and no other firm ever sees them — strict per-firm isolation, no cross-firm training pool, no shared examples. On Enterprise with Private AI, drawing content sent for AI processing never leaves infrastructure your firm controls. Questions from your security team? We're happy to walk through the full data-flow — [email protected].

Native Revit add-in

BIM Monkey installs as a trusted Revit add-in — not a browser extension, screen recorder, or external application with clipboard access. It operates only within the Revit process and communicates exclusively through the Banana Chat panel you open.

Your Anthropic API key stays local

On the Standard and Pro plans, you enter your Anthropic API key during installation. It is saved to a local configuration file on your machine (~/.claude/settings.json). BIM Monkey does not transmit or store your Anthropic API key on its servers. On the Enterprise plan with Private AI enabled, workstations don't use an Anthropic API key at all — inference routes to your own cloud account.

Subprocessors

We share data with a small set of processors solely to operate the service. The principal ones:

Anthropic, PBC	AI model inference (drawing analysis, generation, Banana Chat) — United States. Not used for inference when Private AI is active — Enterprise firms' inference runs in their own AWS or Google Cloud tenancy.
Railway Corp.	API hosting and database infrastructure — United States
Netlify, Inc.	Web application hosting — United States
Amazon Web Services	File/image storage (S3) — United States
Stripe, Inc.	Payment processing — United States
Clerk, Inc.	Authentication and identity — United States
Loops	Transactional email — United States

The complete list, including optional integrations (Autodesk, Google Drive, Dropbox, OneDrive, site-data providers), is maintained in our Privacy Policy.

Data deletion

You can archive or delete individual projects from the dashboard at any time. To delete your entire account and all associated data, email [email protected] — deletion requests are honored within 30 days. Firms using customer-hosted data residency additionally control their own database directly and can revoke BIM Monkey's access at any time.

Vulnerability disclosure

If you believe you've found a security vulnerability in BIM Monkey, report it to [email protected] with the affected system, reproduction steps, and potential impact. We aim to acknowledge valid reports within five business days. Please do not access other customers' data or disrupt the service while investigating.

Questions?

Email us at [email protected]. We're happy to answer IT security questions, discuss enterprise requirements, complete security questionnaires, or provide documentation for your firm's security review process.