On this page
Dranz is a coding-agent workbench that names the provider and privacy tier on every session. Dranz has one route: every session runs on Venice's private tier. Dranz Pro adds local models, direct providers and other routes, and it is in a private build. This page describes those routes too, so you know what Pro changes.
This page explains the current data practices for dranz.ai and the Dranz and Dranz Pro desktop apps. It is a practical notice, not a promise that a provider or tool will never retain or disclose data. The route, provider terms, approved tools, device and your settings all matter.
What we collect directly
Website signup
The signup form on this site collects the email address you submit and the time it was added. We do not ask for a name or phone number, and we do not use the address for a newsletter or a follow-up sequence.
Early access. You get one email confirming the signup and one email on the day a Dranz build is published. Nothing between the two, and nothing after.
The address is stored in our Upstash Redis account, accessed through the site's serverless deployment, and the follow-up email is sent through Resend.
You can ask us to remove or correct a signup address by emailing privacy@dranz.ai from that address. We keep it until the follow-up is complete, you ask us to delete it, or a longer period is required for legal, security or record-keeping reasons. Operational backups and provider logs may take longer to clear.
The site also receives ordinary technical information needed to deliver and protect a web request, such as request time, URL, browser and device information, and network information in hosting or security logs. We use it to operate the site, prevent abuse, debug failures and understand aggregate traffic.
The site does not use advertising cookies or an account login. Vercel Web Analytics is enabled for aggregate page-view measurement; Vercel says its Web Analytics uses anonymized data and does not use cookies. Vercel Speed Insights measures performance and Core Web Vitals. These are third-party services and their own privacy terms apply.
Separately, the site counts a small set of named actions itself: download clicks, product-demo interactions, views of the Dranz Pro section, clicks on the Pro email link, signup and feedback submissions, blog reads, GitHub link clicks and install-guide views. Each is stored as a daily total in our Upstash Redis account, sometimes with a short label such as which demo or which download. No identifier, IP address, user agent or cookie is stored with these counts, and two actions by the same person are simply two counts. Browsers with Do Not Track turned on are not counted at all.
App and support
The app stores the working state needed by the workbench on your device: settings, provider configuration, session and recovery metadata, cost records, command history and drafts. The underlying agent engine may retain its own transcripts. Provider credentials remain on the device and are not sent to the browser.
If you contact us, we receive the information you choose to include, such as your email address, app version, screenshots, logs and the description of a problem. Do not send secrets or sensitive project content in a support request.
Feedback submissions
The feedback form collects the category and note you submit, and the optional Dranz version, engine, system, model, expected outcome and follow-up email. The text is stored in our Upstash Redis account for human review. The form does not upload files and does not send your submission to a model or automatic summarizer. If you email a screenshot or log to hello@dranz.ai, that message and attachment are handled through our email system instead.
Do not include passwords, API keys, private keys, confidential client data or other sensitive information. We use feedback to understand problems, respond when you ask for follow-up, and improve Dranz. We may retain submissions as long as reasonably needed for those purposes, security and record-keeping; ask privacy@dranz.ai if you need a submission removed.
Dranz is not your inference provider. Data handling follows the route you enable: Venice receives the account and operational metadata it needs, Venice can separate your identity from an upstream provider, and in Dranz Pro a local model involves no inference provider at all. The session badge makes that route explicit before work begins.
Routes in Dranz
Dranz has one inference destination. Every engine (Claude Code, OpenCode, Codex, Kimi Code, Grok Build) runs its loop on your machine and sends its model calls to Venice under your Venice key. No other provider account is involved.
| Route | Identity and operational data | Prompt handling |
|---|---|---|
| Venice · private | Venice receives the account and operational metadata required for billing, abuse prevention and reliability. | Venice provides a contract-enforced zero-retention tier: prompt and response content is processed for inference and not retained after the request completes. |
| Venice · anonymized | Venice routes the request while withholding your identity from the upstream provider. | The upstream provider may see the prompt. This protects identity; it is not a zero-retention guarantee. |
Routes in Dranz Pro
Dranz Pro adds the routes below. They are not in Dranz, and Pro is in a private build. This is what Pro changes about where a session goes.
| Route (Dranz Pro) | Identity and operational data | Prompt handling |
|---|---|---|
| Local: Ollama or LM Studio | No cloud account is involved in inference. | The prompt is processed by the model service on this machine. |
| Claude direct | Anthropic receives the account associated with your Claude subscription. | Anthropic's terms and retention policy apply. |
| Coding plans and compatible providers | The provider receives the configured account and operational data. | That provider's terms apply. Dranz Pro does not label these routes private. |
In Pro, a local model keeps inference on your machine when the model is capable enough for the work; a direct provider is for a particular model or an account you already trust. Venice private remains the first hosted route in both products.
How to read the session badge
Provider choice is fixed when a session starts. Dranz shows the route that actually reached the model and does not silently move a session to another provider when the selected route is unavailable.
| Label | What it means |
|---|---|
| Local (Dranz Pro) | Inference is served by Ollama or LM Studio through this machine's loopback connection. |
| E2EE + TEE | Selected Venice models can encrypt tool-free turns after the client checks the available attestation evidence. |
| Private | Venice's contractual zero-retention tier. This is a provider commitment, not an end-to-end encryption claim. |
| Anonymized | Venice withholds your identity from the upstream provider, which may still receive the prompt. |
Encryption and tool use
Venice's encrypted route does not return the structured tool calls a coding agent needs. On a supported model, tool-free turns can use E2EE + TEE; turns that require tools use Venice private and its contractual zero-retention guarantee. Dranz reports the route that was used instead of presenting the stronger label when it did not apply.
Attestation is useful but bounded evidence. Dranz does not currently verify the quote signature against a published reference measurement, so the product does not present attestation as an independent cryptographic proof.
Network access beyond inference
Model inference is only one form of network access. Content may also leave the device when you approve or enable:
- WebFetch, WebSearch, MCP servers or network-capable shell commands;
- a provider's account, billing, abuse-prevention or operational systems;
- the optional model gallery, which can contact a model registry when opened;
- the packaged app's update check; or
- the website's aggregate, cookieless analytics, action counts and performance measurement.
Those actions are not covered by Venice's inference-retention commitment. Dranz makes network-capable actions visible and asks for approval where the workbench can intervene. A local session means model inference is local; it does not automatically make every approved tool action offline.
We do not sell your prompts, code or provider credentials. We do not use private app content to train a general-purpose model. However, the selected provider, tool, registry or account system may process the content you send to it under its own terms. If you use a direct provider (Dranz Pro) or a network-capable tool, that service is a separate recipient.
Data stored on your device
Dranz stores the working state required by the workbench: settings, provider configuration, session and recovery metadata, cost records, command history and drafts. The underlying agent engine may also keep its own transcripts.
Use Settings › Start fresh to remove local history. Start fresh preserves credentials, logins and engine settings so you can clear work without disconnecting the app.
Provider credentials remain on the device and are not sent to the browser. Never paste a credential into a prompt, issue, social post or support reply.
Assists and automations
Session headlines follow the provider for the active session. Other assists use the providers you configure for that purpose, in the order shown in Settings. If the selected provider cannot start, the assist stops and explains the problem; it does not quietly send the content elsewhere.
If content should not leave the device, keep the session local and avoid network-capable tools. The provider badge identifies the inference route, and your approvals determine which additional actions run.
Lookout, the optional second opinion, sends excerpts of a session to a separate Venice tier when you turn it on. Lookout lists exactly what it sends.
Your choices and privacy rights
You can pick the model per session, decline a provider, disable network-capable tools, skip the signup list, or ask us to delete information we hold. In Dranz Pro you can also choose a local route. Depending on where you live, you may also have rights to access, correct, delete or receive a copy of personal information, object to or restrict certain uses, or complain to a privacy regulator. Send a request to privacy@dranz.ai. We may need to verify the request and may retain a limited record of it to comply with law and prevent abuse.
We do not currently sell or share personal information for cross-context behavioral advertising. If that changes, this page and any legally required choice mechanism will be updated before the change takes effect.
Security and changes
We use reasonable administrative, technical and organizational safeguards, but no device, network, provider or software system is perfectly secure. You are responsible for backups, credential hygiene, permissions, and choosing a route appropriate for the sensitivity of your work.
We may update this page as the products change. The effective date above will change with a material update. If a future feature changes what we collect or where content goes, we will update the relevant disclosure before or when the feature is introduced.
Quick reference
Dranz runs every session on Venice's private tier, a hosted route with a zero-retention contract. Dranz Pro adds local models for when inference must remain on your machine, and direct providers for when a particular model or subscription is the right fit. Dranz keeps the route visible; the guarantee belongs to the provider you select.
For the operator's contract terms, risk allocation and early-software disclaimer, read the Terms of Use. For practical safeguards, read Safety & responsible use.