What we hold, and why
Privacy notice.
Murmello is a Mac app that runs coding agents in the cloud and on your Mac, with hosted dictation. Some of it keeps to your Mac. When you use an agent, Murmello saves versions of its work, including the contents of a folder you let it use on your Mac, in our storage; a cloud agent also runs on a machine we rent from Fly.io; and your dictation audio goes to a transcription provider. This page says exactly which is which.
Effective October 31, 2026 · Tomas Godoy Pastore · support@murmello.ai
The short version
Nothing leaves your Mac until you sign in and use an account feature. When an agent works, versions of its workspace (its files, git history, its tool's history and the conversation) are saved in our storage, and that includes the contents of a folder on your Mac that you let an agent use. A cloud agent runs on a machine at Fly.io. Sign-ins you keep only on your Mac never leave it. Your dictation audio is sent to ElevenLabs to be turned into text. We do not sell your data, we do not use your files or audio to train models, and we do not show you advertising.
1. Who is responsible
Tomas Godoy Pastore, an individual doing business as “Murmello” in the State of Florida, United States, is the controller of the personal data described here. For any privacy question or request, write to support@murmello.ai. We are a one-person operation and have not appointed a data protection officer or an EU/UK representative; that address reaches us directly.
2. What stays on your Mac
Your microphone audio is captured locally and streamed only while you are dictating. Idle audio captured by the optional Instant Start feature is discarded without being sent anywhere. Accessibility permission is used to identify the target field and paste; that inspection happens locally and is not uploaded. Settings such as your shortcut, microphone choice, language and appearance are stored in macOS preferences on your device. Your work in the app's terminals, files in folders you have not let an agent use, and sign-ins you keep only on this Mac stay on your Mac. When you let an agent use a folder, its full path on your Mac is not sent to us, only its name.
Murmello places the recognised text on your clipboard and can paste it into the active app. Once text is pasted elsewhere, the receiving app's own data handling applies — it may sync your text to its servers, and that is outside our control. macOS features such as Universal Clipboard behave the same way.
3. What we collect, and why
Account information
When you create an account we receive your email address and basic profile information from your sign-in provider (Google, GitHub or Apple), or your email address and a password hash if you register directly. We use it to identify you, secure your account, send service email such as verification and password resets, and contact you about your subscription. Legal basis: performance of our contract with you.
Your agents and their work
We keep the profiles you make (a coding tool, its instructions, model,
permissions, skills, connections and settings), the teams and boards you
save, a record of each agent (its name, state and where it runs), and the
messages between you and your agents, and between agents you connect,
which pass through our servers.
As an agent works, we save versions of its workspace: at the end of each
reply, every ten minutes while a reply runs, when it waits for your answer,
and before it sleeps. A version holds the agent's files, each repository's
git history, its coding tool's own history and memory, and the
conversation. Caches and installed dependencies are left out; files that
git ignores, such as .env, are kept; sign-in files are never
saved. Large files are stored once for your account. We also keep caches
of the tools and packages your agents' environments install, for your
account only, so later agents start faster. Legal basis: performance of
our contract.
Agents on your Mac
An agent on your Mac runs in a virtual machine on your Mac. It works in a folder you grant or in a copy we keep for it, and versions of that folder's contents are saved in our storage as described above, so the agent can move to the cloud, be restored or be forked. For each Mac you set up we keep its name, macOS version, chip type and helper version, whether it is online, whether its disk is nearly full, and the names of the folders you granted. Your Mac's public IP address, which our servers see for any connection, is also used to stop agents from reaching your own network. The agent's own internet traffic goes from your Mac directly to where it is going, not through us. Legal basis: performance of our contract.
Dictation audio
While you dictate, 16 kHz mono audio is streamed to our server and forwarded to ElevenLabs, which returns text. We do not store your audio or your transcripts: audio is held in memory only for the length of the utterance, and the resulting text is returned to your Mac and not retained by us. We ask ElevenLabs not to log the content of these requests. We do record the number of seconds of audio processed, because that is what your allowance is measured in. Legal basis: performance of our contract.
Sign-ins for your coding tools, and GitHub
When you add a sign-in for a coding tool, you choose where it is kept. In Murmello's vault, it is encrypted with a key from AWS Key Management Service, bound cryptographically to your account, and there is no endpoint that reads it back out; it is handed to an agent's machine only while that agent runs, and a renewed login is saved back to the vault. Only on this Mac, it is kept in your Mac's Keychain: we store only a reference to it, it is handed to agents on that Mac without passing through us, and agents elsewhere cannot use it. You can remove a stored sign-in at any time, and it is deleted.
If you connect GitHub, we store its access token encrypted in the same way. It is used only on our servers, for the imports, pulls and pushes you ask for, and never handed to an agent. Disconnecting revokes it at GitHub. Legal basis: performance of our contract.
Skills and connections you add
You choose the skills and connections your agents use. When you add a skill, we keep its files in your library, fetched from the source you picked or as you uploaded them. A connection is an MCP server run by whoever provides it, not by us: when an agent uses it, the agent's tool calls and their results go to that server, with any token or key you give it, under that provider's own terms. If you sign in to a connection, we keep its token encrypted in the vault, as we keep your other sign-ins. Legal basis: performance of our contract.
Payment information
Stripe processes your payment and holds your card details. We receive and store your Stripe customer and subscription identifiers, your plan, your invoice and payment status, and the country and tax status needed to bill you. We never see or store your full card number. Legal basis: performance of our contract and compliance with tax and accounting law.
Usage and operational records
We record what you consume against your allowance — the cloud credits your cloud agents use, measured by how long each one runs, and dictation seconds — together with reservation and settlement records so billing is auditable. We also keep short-lived request counts to enforce rate limits. Our servers keep operational logs containing IP addresses, timestamps, request paths and error information, used to run and secure the service and to diagnose faults. Legal basis: performance of our contract and our legitimate interest in operating a secure, working service.
This website
The marketing site is static, hosted on Cloudflare, and carries no advertising, analytics scripts, third-party fonts or tracking cookies. Cloudflare processes connection information needed to serve and secure it. The billing page is served by our own server and uses a session cookie that exists only to keep you signed in; it is strictly necessary and is not used for tracking. We look at the aggregate request counts Cloudflare keeps for the site — how many visits, which pages, which countries — to see whether it is reaching people. Those counts come from serving the site, not from a script or cookie, and are not tied to you.
The early-access waitlist
If you join the waitlist on murmello.ai we keep your email address, when you joined, and — if the link you arrived by carried one — a short tag naming the post or advert it came from, so we can tell which ones reach people. Nothing else is collected, and joining creates no account. We use your address only to email you about early access: your invitation, and news about the launch. We do not sell it, share it for advertising, or add you to anyone else's list. Legal basis: your consent, which you give by joining and can withdraw at any time — every email we send has an unsubscribe link, or write to support@murmello.ai and we will remove you. The list is stored on our infrastructure in AWS, and emails are sent through Amazon Simple Email Service.
4. What we do not do
We do not sell your personal information or share it for cross-context behavioural advertising. We do not use your files, your audio or your transcripts to train machine-learning models, and our agreements with our providers are configured to prevent them doing so on our behalf. We do not read your project files except where you ask us to investigate a specific problem, or where we are legally compelled. There is no advertising anywhere in Murmello.
5. Who else processes your data
We use a small number of providers to run the service. Each one is listed, with what it handles and where, on our subprocessors page. We will update that page before adding a new one that handles your content.
We may also disclose data if we are legally required to, or where we genuinely believe it is necessary to protect our rights, your safety or the safety of others. If Murmello is ever incorporated or its business transferred, data may transfer with it under this notice; we will tell you first.
6. Where your data is processed
Our infrastructure runs in Amazon Web Services' us-east-1 region
in the United States, and your account data, the saved versions of your
agents' work and billing records are processed there. A cloud agent runs on
a machine at Fly.io in the United States, placed near our server in
Virginia, which creates it. If you are in the EEA, the UK or Switzerland, this means
your data is transferred outside your home jurisdiction. We rely on the
European Commission's Standard Contractual Clauses (and the UK Addendum where
relevant) as the transfer mechanism with our providers.
7. How long we keep things
Account data, profiles and agent records are kept while your account exists. Saved versions of your agents' work are kept while the agent exists: a cloud agent asleep for 30 days is archived, its cloud machine is removed, and every version is kept so Restore can bring it back. When you end an agent, its cloud machine is removed and its saved versions are deleted 30 days later. Two things are kept longer: versions that another of your agents still builds on (a fork made from it) are kept while that fork exists, and large files (over 16 MiB), which your account's agents share, are kept until your account is deleted. Nothing else deletes versions automatically. Our storage keeps replaced or deleted copies for up to 30 days, after which they are gone. Caches of your agents' installed tools expire on their own within 60 days. Dictation audio is not retained at all, and transcripts are not stored by us. Usage and billing records are kept for as long as we need them for accounting and tax purposes, typically seven years. Operational logs are kept for a short period for security and debugging. A waitlist address is kept until you unsubscribe or ask us to remove it, or until we close the waitlist, whichever comes first.
To delete your account, write to support@murmello.ai. When your account is deleted, all of its stored agent files are deleted: every saved version, the large files kept once for your account, and leftovers from transfers; their backup copies are erased within 30 days. Database records go with the account, except records we must keep for legal or accounting reasons, such as billing. Disconnecting a Mac, with “Also remove the files”, removes the virtual machines and their files from that Mac. Files on your own Mac, including the folders you granted, are never deleted by cloud cleanup.
8. Security
Connections use TLS. Saved versions and the database are encrypted at rest. Sign-ins in the vault and the GitHub token are encrypted with per-write keys from AWS Key Management Service and bound to your account. Access to production systems is limited to the operator and uses scoped roles rather than shared long-lived keys. Each cloud agent runs on its own machine, and each agent on your Mac in its own virtual machine, which sees only the folder you granted and cannot reach your local network.
No service is perfectly secure, and we will not pretend otherwise. If a breach affects your personal data we will notify you and the relevant authority as the law requires. If you find a vulnerability, please report it to support@murmello.ai before disclosing it publicly.
9. Your rights
Depending on where you live, you may have the right to access the personal data we hold about you, to correct it, to delete it, to receive a portable copy, to object to or restrict certain processing, and to withdraw consent where we rely on it. To exercise any of these, write to support@murmello.ai. We will respond within the time the law allows — 30 days in most cases — and will not charge you or treat you differently for asking.
If you are in the EEA or UK, you may also lodge a complaint with your local supervisory authority. If you are in California, you have the rights to know, delete, correct and opt out of sale or sharing under the CCPA as amended by the CPRA; we do not sell or share personal information, and we do not use or disclose sensitive personal information for purposes requiring an opt-out.
10. Children
Murmello's services are not for people under 18, and accounts require you to be 18 or older. We do not knowingly collect personal data from children. If you believe a child has given us data, write to support@murmello.ai and we will delete it.
11. Software updates and downloads
Murmello checks for signed updates using Sparkle. Checks contact https://murmello.ai through Cloudflare, and update downloads come from Amazon CloudFront. When you set up your Mac for agents, it downloads a Linux image once from the same place, checked against fingerprints built into the app. These providers receive ordinary connection information, including your IP address, the requested URL and a user-agent string that can contain app and system version information. Sparkle system profiling is disabled. Update requests contain no recordings, transcripts, account identifiers or usage analytics, and you can turn automatic checks off in Settings → Updates.
12. Changes to this notice
We will update this page when what we do changes. If a change materially affects how we handle your personal data, we will tell you by email or in the app before it takes effect. The effective date at the top always tells you which version is current.
Contact
Privacy questions, requests and complaints go to support@murmello.ai, and reach Tomas Godoy Pastore directly.