Privacy Policy

AgentMug is operated by Dahshan Labs. This policy explains how we handle information when you use agentmug.com, the hosted AgentMug API, or a connected AgentMug worker.

Portable workers can also run locally or in a self-hosted environment. Information processed entirely there is controlled by that environment's operator. This policy applies when a worker communicates with an AgentMug-hosted account, connection, or service.

• Account data: account identifier, email, name, profile image, authentication state, and support communications.
• Worker content: instructions, prompts, blueprints, schedules, documents, conversations, run inputs and outputs, tool results, traces, evaluations, and feedback that you choose to provide.
• Connection data: provider, connected-account label and identifier, granted scopes, encrypted OAuth tokens or credentials, status, and identifiers returned by actions performed for you.
• Service data: plan, usage, token and cost measurements, billing status, request path, timestamps, reliability, security events, and limited device information.

When you connect Facebook, Instagram, or Threads, Meta presents its own authorization screen. AgentMug never asks for or stores your Facebook or Instagram password. Depending on the permissions you approve and features available to your account, AgentMug may receive:

• your Meta user identifier and basic connected-account information;
• Facebook Pages you manage and their identifiers, names, and Page access tokens;
• linked Instagram professional-account identifiers and profile information;
• Threads profile identifiers and publishing permissions;
• content you ask a worker to publish, plus returned post, media-container, and status IDs;
• comments or messages only where an approved AgentMug feature uses them and you explicitly grant the required permission.

We use this data to show the accounts you can select, execute the publishing or management action you requested, report its result, maintain the connection, prevent abuse, and provide support. We do not sell Meta Platform data, use it for behavioral advertising, or use it to train a general-purpose AI model.

AgentMug accesses Google account data only after you choose to connect an account and approve permissions on Google's consent screen. The data available depends on the scopes you grant and the feature you use; connecting one Google service does not grant access to every Google service.

• Account identity (userinfo.email): we read your connected email address to label the connection and help you choose the correct account.
• Gmail: with gmail.readonly, a worker can search and read message and thread identifiers, sender, subject, date, snippets, and unread status for your requested inbox summaries or triage. The current reader does not retrieve full message bodies or attachments. With gmail.send, it can send the recipient, subject, and plain-text body supplied for your requested email action. A draft feature uses gmail.compose only when that additional permission is granted, and stores the proposed message as a Gmail draft.
• Google Calendar: granted calendar permissions let a worker list event identifiers, titles, times, descriptions, locations, attendees, and event links from your primary calendar to answer your scheduling requests, or create an event from the details you provide. The calendar.events permission is used for event creation.
• Google Sheets (spreadsheets): a worker sends the values you request to the spreadsheet and range you identify, and receives the updated range and row count so it can report the result.

YouTube is currently limited to integration testing. Test authorization requests youtube.readonly for channel identity. This permission does not establish that channel access has been verified. Worker video uploading is not currently available in AgentMug.

We use Google data to identify the connected account, perform the feature you request, show its result, maintain the connection, and provide support and security. OAuth access and refresh tokens are stored encrypted on the server; YouTube test tokens are stored in AgentMug's self-hosted Nango service. We also store the account label, granted scopes, and connection status. Tokens are not included in prompts sent to AI models.

When your requested worker task uses AI, relevant Google content and tool results may be sent to the model provider selected for that run, such as Anthropic or Google, to produce the user-facing result. Message snippets, event details, spreadsheet values, generated drafts or summaries, and action results may also appear in your conversations, run history, and execution traces. Hosting and connection providers process the data needed to operate these features, as described in the disclosure section below.

AgentMug does not use data obtained through Google Workspace APIs to develop, improve, or train non-personalized AI or machine-learning models. We do not sell Google user data or use it for advertising or credit decisions.

AgentMug's use and transfer of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements. Human access to Google user data is limited to your affirmative permission for specific data, necessary security or abuse investigations, legal requirements, or aggregated and anonymized internal operations permitted by that policy.

Disconnecting a native Google connection in AgentMug removes its stored credential and worker bindings. You can separately revoke Google's authorization in your Google Account connections. Revoking access at Google does not by itself delete AgentMug run records or content already sent to a connected service. Google-derived conversations, tool results, and traces follow the retention rules below. To request deletion of that content or a test connection that is not shown in AgentMug, email privacy@agentmug.com from your account address and identify the connection or records to remove.

• authenticate you and operate, secure, and support AgentMug;
• build, run, schedule, monitor, export, import, and improve your workers;
• execute actions you request through connected services;
• provide history, traces, memory, evaluations, and connection status;
• measure usage, enforce plans, process billing, and investigate failures or abuse;
• comply with law and enforce our Terms of Service.

To complete a requested run, AgentMug can send the prompt, relevant worker instructions, conversation context, tool results, and only the connected-service data needed for that run to the selected AI model provider. Current hosted deployments can use Anthropic or Google.

A user may review a generated draft or explicitly configure a scheduled or automated action. You control which accounts a worker can use and can disconnect them at any time. Private worker content and connected-service data are not used to train AgentMug's own general-purpose model.

We disclose information only as needed to provide the service, including to:

• providers you connect, including Meta, Google, Microsoft, LinkedIn, X, TikTok, and other services, to perform your requested action;
• operational vendors such as Clerk, Railway, Nango, Anthropic, Google, Stripe, Twilio, and Resend, where the corresponding feature is enabled;
• other people when you intentionally publish or publicly share worker content;
• authorities or affected parties when reasonably required by law, safety, or security;
• a successor in a merger, acquisition, financing, or sale, subject to applicable law.

We do not sell personal information or connected-service data to data brokers.

Connected-service credentials are stored server-side and encrypted at rest. Nango-backed OAuth connections use AgentMug's self-hosted Nango service for token storage and refresh. Page tokens and provider secrets are not exposed to workers or browsers after authorization.

We use HTTPS, access controls, tenant scoping, audit and security logging, and authenticated encryption. No security measure is perfect, so protect your account and contact security@agentmug.com if you suspect unauthorized access.

• Connection credentials remain until removed through disconnection or a deletion request. Revoking access at a provider does not by itself delete AgentMug records.
• Worker, conversation, and run records remain while needed to provide your account, subject to product history limits, security, billing, backup, dispute, and legal requirements.
• Optional raw product-analytics events are retained for up to 180 days.
• Deletion from active systems may be followed by deletion from backups on their normal lifecycle, unless retention is legally required.

See our Data Deletion Instructions to disconnect a provider or request deletion of your hosted data.

You can manage or disconnect providers in AgentMug's Integrations settings and revoke Meta access in Facebook's Business Integrations settings. Depending on your location, you may also have rights to access, correct, export, delete, or object to processing.

Send requests from the email associated with your account to privacy@agentmug.com. We may verify your identity before acting and will respond within the period required by applicable law.

Essential cookies and storage support authentication, security, navigation, and preferences. Optional first-party product analytics run only after consent and exclude raw prompts, model output, credentials, form values, and full referrer URLs. AgentMug honors Global Privacy Control as a denial of optional analytics.

AgentMug and its providers may process information in countries with different data laws. We use legally required safeguards for applicable transfers.

AgentMug is not directed to children under 18, and we do not knowingly collect their personal information. Contact us if you believe a child has provided information.

We may update this policy as AgentMug changes. We will revise the effective date and provide additional notice when required.

Contact Dahshan Labs at privacy@agentmug.com for privacy questions, access requests, or deletion requests.