Your infrastructure.
Your choices, explained.
MPLOI Server is self-hosted software. Where information is processed also depends on the models, tools and connected services you choose. This policy explains those boundaries, the information mploi.ai handles, and what disconnecting or deleting data does — and does not — remove.
Last updated
Privacy at a glance
Self-hosted storage does not mean that no data leaves your server.
Your instance. Accounts, conversations, documents, credentials and agent records are held on the infrastructure running your instance. The deploying organization administers that infrastructure and its access, retention and backup arrangements.
Your selected destinations. An authorized connector, online tool, browser action or cloud-model request can send information to the relevant service. OAuth and optional event relays can also involve mploi.ai. Support information you choose to send us and visits to this website are separate from your instance's stored data.
Optional capabilities, not automatic access. Integrations described here depend on availability in your release, administrator settings, user authorization and tool permissions. Listing an integration in this policy does not enable it, connect an account or grant consent.
Who is responsible for what?
For a customer-operated instance, the organization deploying MPLOI normally determines why and how instance data is used and acts as the data controller, or the equivalent role under applicable privacy law. It is responsible for informing its users, establishing an appropriate basis for processing, managing permissions and connected accounts, choosing service providers, securing the deployment and responding to requests about data it holds.
Installing self-hosted software does not itself give the MPLOI team routine access to your instance database. However, information may reach us through the website, the relays described below, or support access or material you voluntarily provide. This policy covers those activities as well as explaining instance processing; it does not replace the deploying organization's own privacy notice or the policies of connected services.
Use of MPLOI is also addressed in our Terms of Service.
What your instance holds
The categories depend on the features you use and your deployment settings.
Accounts and access. Usernames, email addresses, group memberships, authentication records and identity information supplied by a sign-in provider.
Conversations and agent work. Messages, responses, uploaded documents, extracted content, tool calls and results, browser screenshots, saved memories, schedules and run history. Information read from a connected account can become part of these records and of outputs shared with other people who have access to them.
Credentials and connection metadata. OAuth access and refresh tokens, account identifiers, granted scopes and connection status; separately saved credentials and their access grants; and browser pairing information. Connector tokens are encrypted in the instance database and associated with the user who authorized the connection. Saved secret fields are protected separately from descriptive metadata.
Configuration and operational records. Agent instructions, model and tool settings, sign-ins, permission changes, credential-access events, tool activity and errors. Administrators' permissions and the instance's sharing settings affect who can access records and resulting content.
Microsoft 365 and Microsoft Graph
Available tools act through an authorized user's connection, within the permissions granted.
Where the optional Microsoft 365 connector is available and enabled, it uses Microsoft Graph. Access depends on the account's own rights, the scopes consented to by the user and, where required, the Microsoft tenant administrator, together with the tools permitted for the agent. A connection does not imply access to every Microsoft service or every item in a tenant.
Depending on those permissions and the task, data may include account/profile identifiers; Outlook email subjects, senders, recipients, bodies, folders and attachments; calendar events, attendees and availability; contact names and contact details; OneDrive file and folder metadata and file contents; Teams teams, channels, chats, messages and meeting information; and SharePoint site, document-library and document information.
These tools support requested work such as finding or summarizing information and preparing or carrying out authorized actions. If permitted, actions can include sending, replying to or drafting email, creating or changing calendar events or meetings, posting Teams messages, and uploading, sharing or deleting files. Availability and confirmation requirements vary by tool and configuration. Authorized scheduled jobs may also use a connection within their allowed permissions; access is not necessarily limited to the moment of an interactive chat.
Graph content requests normally run between your instance and Microsoft. OAuth identity, token exchange and refresh can use the mploi.ai relay. Retrieved content may be retained in instance records or included in requests to your selected AI provider as described under model processing. Microsoft also processes information under its own terms and the organization's Microsoft service arrangements.
Google Cloud and Google user data
The optional Google Cloud connector acts through a user's own Google account, within the access that account already has.
When a user connects Google Cloud, MPLOI asks Google for these
permissions: the user's basic profile and email address
(openid, email, profile),
used to show which account is connected, and
https://www.googleapis.com/auth/cloud-platform,
used to manage the user's Google Cloud resources on their
behalf. Google Cloud's own access controls (IAM) still decide
every request: the agent can only see and change what the
user could see and change in the Google Cloud console.
Depending on the user's permissions and the task they ask for, data accessed can include project, folder and organization details; resource configuration (for example Compute Engine, GKE, Cloud Run, Cloud Storage, BigQuery, Cloud SQL and other services); logs, metrics, traces and error reports; IAM policies and service accounts; billing and budget information; and the metadata of Secret Manager secrets. Secret values are never shown to the AI model or stored in conversations. With the user's approval, the agent can make the changes the user asks for, such as creating, changing or deleting resources.
How we use Google user data. MPLOI uses data received from Google APIs only to provide and improve the user-facing features the user asked for, in that user's own MPLOI instance. Google OAuth tokens are stored encrypted in the customer's own MPLOI instance and are used only to make the requests the user's agents need. The mploi.ai OAuth relay handles the sign-in, token exchange and refresh; it does not keep the tokens or call Google Cloud APIs on its own.
MPLOI's use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements. In particular:
- Google user data is not sold, and is not used or transferred for advertising, including personalized, retargeted or interest-based ads.
- Google user data is not used to determine creditworthiness or for lending purposes.
- Google user data is not used to develop, improve or train generalized or non-personalized AI or machine-learning models.
- Google user data is transferred to others only as needed to provide the features the user asked for (for example, the AI model provider the customer selected processes a request's content to answer it, as described under model processing), to comply with law, or as part of a merger or acquisition with notice to users.
- No person at MPLOI reads Google user data unless the user gives explicit permission for specific data (for example in a support request), it is necessary for security purposes or to comply with law, or the data is aggregated and anonymized for internal operations.
Users can disconnect Google Cloud at any time from the chat's connectors menu, which deletes the stored tokens and revokes MPLOI's access with Google. Access can also be removed at myaccount.google.com/permissions. Information retrieved during a task is kept in the customer's instance as described under retention.
Other connected services
Optional integrations can include GitHub, Atlassian services, Azure, Google Cloud, Slack and other configured tools or APIs. Depending on the integration, an authorized task may access repositories, issues, pages, messages, files, cloud-resource information or other records, and may make permitted changes. Credentials, provider permissions, administrator settings and agent/tool permissions determine the available access.
Content API calls generally go from the instance to the connected provider. That is not a promise that all information stays within the instance: API requests and authorized writes are sent to that provider, tool results may be stored or sent to a model, and the OAuth and Slack event relay paths described below are exceptions to a direct-only route. Connected services apply their own privacy, security and retention terms.
Browser Helper
Working in your browser can involve information from sites where you are already signed in.
Browser Helper is an optional browser extension paired with your MPLOI instance. When enabled and authorized, it exchanges commands and results with that instance. These can include tab URLs and titles, page text and interactive elements, screenshots, action details and results, and browser/device or extension information needed to maintain the connection. Screenshots can accompany tool actions as well as explicit screenshot requests.
Authorized actions can navigate, click, type, select or submit information on a page, using the access available in that browser's signed-in session. Where supported and permitted, a sign-in action can send credentials from your instance's credential store to the paired browser to fill the site's login form. You can also sign in yourself. Pairing, site allow/block settings and action approvals govern this feature; they do not make all page content safe to disclose.
Page content, form information or images can reveal personal, confidential or third-party information. Results and screenshots may be saved with the conversation and sent to a selected model that accepts them. Sites receive the requests and submissions made in the browser under their own policies. Share only tabs and actions you are entitled to authorize. Unpairing or revoking browser access stops that connection; it does not erase earlier conversation records, screenshots or actions already completed on a site.
Local models, cloud models and online tools
Local or privately hosted inference. MPLOI can use a model endpoint operated on infrastructure you control. Model requests go to that configured endpoint. Keeping inference within your environment depends on its actual location and configuration, and does not prevent a separately enabled connector, browser or online tool from making external requests.
Cloud inference. If a cloud AI provider is selected, the instance sends the information needed for that request to that provider. This can include prompts and agent instructions, relevant conversation context, document content, tool results and supported images such as browser screenshots. Connected-account content, including Microsoft data, can be included when it is used in the task. Providers may process requests outside your organization or country; their handling, retention and any model-training use depend on the provider, account settings and applicable service agreement, not solely on MPLOI being self-hosted.
Online tools. Search, web-reading, API and other online tools send queries, request metadata or task content to their destinations. Authorized uploads, posts and other actions send the data required to carry them out. Review your enabled tools and providers before using sensitive information.
Data loss prevention is a safeguard, not a boundary guarantee. When configured, DLP can mask detected sensitive text before a model request. Its effectiveness depends on enabled rules, supported content and successful scanning. It may miss data; images are not assured to be redacted; and a failed scan or timed-out rule can leave text unmasked. DLP does not guarantee anonymization, block every external transfer or replace permissions and an appropriate provider choice.
OAuth identity, exchange and refresh
Some connections use a shared sign-in relay at www.mploi.ai/oauth.
The relay provides a registered callback address for supported providers. To complete a connection, it receives authorization flow information such as the provider, requested scopes, return address and state identifiers, exchanges the authorization code for tokens, and can call provider identity or profile endpoints to label the account. Data handled can include access and refresh tokens, token-response metadata, account identifiers, email/profile information or a workspace identity, depending on the provider.
The completed token bundle is placed in a single-use handoff for the instance to collect through an authenticated request. The browser redirect carries an opaque handoff identifier rather than the token bundle. The handoff has a 60-second validity window and is removed from active handoff state when collected. Pending authorization flows have a 30-minute validity window. Expired flow and handoff records are swept during subsequent state-store operations; expiry is not a promise of physical erasure at that exact second.
The instance stores the connection's tokens for later use. For connections using relay-based refresh, it sends a refresh token back to the relay, which exchanges it with the provider and returns the response. Refresh can occur automatically while a connection remains active, including outside an interactive chat. Supported revocation requests can also pass a token through the relay. These exchanges are distinct from the initial 60-second handoff.
This OAuth path is for authorization and account identity, rather than general retrieval of mailbox, document or other task content. It is not the only circumstance in which mploi.ai may receive information. Relay operations generate operational and error records, and the handoff validity window does not establish retention periods for those logs, refresh processing or infrastructure backups.
Optional Slack event relay
If your deployment uses the mploi.ai Slack event relay, Slack sends event payloads to that relay for delivery to a subscribed instance. These payloads can contain message text and workspace, channel, user or event identifiers. This path can therefore carry connected-service content through mploi.ai, not just sign-in information.
Queued events have a 10-minute validity window and are removed from the active queue when collected. Expired entries are swept during state-store operations. Subscription records link a workspace and an instance for delivery and can remain while registered. These queue rules do not delete messages already delivered to an instance, copies in Slack, operational logs or backups.
Retention, disconnecting and security
Stopping future access and deleting existing records are different actions.
Instance retention. The deploying organization determines retention and backup practices for its accounts, chats, documents, tool results, memories and logs. Some features have specific controls, such as run-history retention or time-limited and single-use credential access. Separately saved credentials can remain until removed, and OAuth tokens may be renewed while a connection is active. No single software-wide retention period applies to all of this data.
Disconnecting. Disconnecting a connector removes its stored active connection credentials from the instance and may attempt provider-side revocation where supported. Revocation is provider-dependent and can fail; you may also need to revoke access in the provider's account or tenant settings. Disconnecting does not remove previously retrieved content in chats, documents, saved memories, outputs, audit records or backups, and does not undo completed actions.
Deletion and other copies. Deletion from an active system does not necessarily erase backups, caches, security logs, legally required records or copies already sent to a provider or shared with another person. Those copies follow the relevant operator's retention and deletion arrangements. Website, relay operational and support records are retained according to their purpose, operational or security needs and applicable requirements; contact us for information about a particular record or deletion request.
Security controls. MPLOI includes controls such as encrypted stored connector tokens, credential access grants and authenticated relay requests. Effective protection also depends on the instance's deployment, administrator choices, browser and provider settings, and the handling of logs and backups. These controls reduce risk; they are not a guarantee that every deployment is secure or that information can never be disclosed.
This website and voluntary support
Contact and demo requests. If you contact us, we receive the details you submit, such as your name, email address, organization and message. We use them to reply, arrange a demo or follow up about your request and MPLOI. Communications and hosting services may process information needed to provide those functions.
Support. Information you voluntarily send for assistance — for example messages, screenshots, diagnostic logs, configuration extracts or example documents — is available to us for investigating and responding to the issue. Any access you arrange to your deployment is separate from ordinary self-hosted operation. Remove secrets and unnecessary personal or third-party information before submitting material, and provide only what you are entitled to share.
Access and security records. The website and its hosting infrastructure may record IP addresses, requested URLs, timestamps, browser user agents and related request or error information to operate the site, diagnose problems and protect against misuse. Such records are separate from your instance's chat database.
Analytics and external resources. This website loads Google Analytics to measure visits and site usage. It can use cookies and similar identifiers and send usage, browser/device and network information to Google. The site also loads Google Fonts and Font Awesome resources from external hosts. Those requests disclose information needed to serve them, such as your IP address and browser information, and may include a referrer subject to browser and site settings. Google's handling is described in its Privacy Policy; other external hosts apply their own policies.
Cookies and browser settings. Cookies or similar storage may also support functionality, session state and security where those features are used. You can manage cookies and tracking through your browser settings or privacy tools; blocking them may affect site functionality or analytics. This page does not itself enable integrations or change the consent and permission settings of a customer instance or third-party service.
Your rights, and who to ask
Depending on applicable law, you may have rights to request access, correction, export or deletion of personal information, to object to or restrict certain processing, or to withdraw consent where processing relies on it. Rights and exceptions depend on the circumstances; withdrawal does not undo earlier lawful processing or completed tool actions. You may also have the right to contact a relevant privacy regulator.
For information held in a customer-operated instance, contact the organization running it. Its administrators control those records and the relevant retention and backup settings. For copies held by a connected service or AI provider, its own request processes and your organization's agreements may also apply.
For website information, relay or support information handled by MPLOI, or questions about this policy, write to info@mploi.ai. Tell us enough to identify the relevant request without sending credentials or unnecessary sensitive information. We may need to verify your identity or authority and will handle applicable rights requests within the period required by law.
Material changes to this policy will be posted here with an updated date. Your deployment's configuration, connected provider policies and the deploying organization's own notice should also be reviewed when features or data uses change.
Questions about
how data is handled?
Contact our team to discuss MPLOI's data flows and your deployment choices.
Talk to Our Team