Terms Privacy
Privacy Policy | External Services & Data Access

External Services and Data Access

This page is part of our Privacy Policy. It lists every external service Memora connects to, the exact permissions we request, the data we retrieve, and what we never access.

1. Our principles

  • No model training. Your workspace data is never used to train public AI models.
  • Least privilege. We request the minimum read permissions needed to index your content.
  • Gated write actions. Actions such as creating a Jira ticket or a GitHub issue require explicit user approval.
  • EU hosting. Indexing and vector storage run on European infrastructure under GDPR.

2. Connected services

2.1 Communication

Slack

Enables Memora to index channel discussions, capture decision threads, and reply with cited facts when mentioned.

Authentication: OAuth 2.0. Access level: Read + Confirmed Action. Sync method: Webhooks + Scheduled Sync.

Permissions (scopes) requested:

  • channels:history
  • groups:history
  • im:history
  • mpim:history
  • channels:read
  • users:read
  • users:read.email
  • app_mentions:read
  • chat:write

Data we retrieve and index:

  • Public & invited private channel messages
  • Thread replies & conversation context
  • User display names and workspace email addresses
  • Shared message attachments & text snippets

Data we never access:

  • Direct messages where Memora is not a participant
  • Channels the bot has not been explicitly invited to
  • Slack workspace billing, credit cards, or admin passwords

Actions Memora can take (only after approval):

  • posttochannel (only upon interactive user confirmation)
  • replyinthread (responding directly to @Memora mentions)

Microsoft Teams

Indexes Microsoft 365 channel history and enables conversational memory responses within Teams channels.

Authentication: OAuth 2.0. Access level: Read + Confirmed Action. Sync method: Webhooks + Scheduled Sync.

Permissions (scopes) requested:

  • ChannelMessage.Read.All
  • Team.ReadBasic.All
  • Channel.ReadBasic.All
  • User.Read

Data we retrieve and index:

  • Channel conversations and thread replies in joined teams
  • Team structure, channel metadata, and member email links
  • Message timestamps and associated mentions

Data we never access:

  • Personal 1-on-1 Teams chats outside Memora bot conversations
  • Corporate Microsoft 365 security logs or Azure directory secrets
  • Audio/video streams from live meetings

Actions Memora can take (only after approval):

  • posttochannel (requires explicit run command)
  • replyToActivity (replying in thread with citation cards)

Gmail

Allows workspace members to ground memory in past email discussions and prepare draft replies without auto-sending.

Authentication: OAuth 2.0. Access level: Read + Confirmed Action. Sync method: Periodic Polling.

Permissions (scopes) requested:

  • https://www.googleapis.com/auth/gmail.readonly
  • https://www.googleapis.com/auth/gmail.compose

Data we retrieve and index:

  • Email message bodies, subjects, senders, and recipients
  • Thread headers, labels, and message dates
  • Plaintext content of email discussions

Data we never access:

  • Memora NEVER automatically sends emails without review
  • Password reset or sensitive verification emails
  • Account settings or Google security credentials

Actions Memora can take (only after approval):

  • draft_gmail (creates a draft in your Gmail inbox for you to inspect; NEVER sends autonomously)

Microsoft Outlook

Reads company emails to build contextual knowledge across projects, commitments, and vendor correspondence.

Authentication: OAuth 2.0. Access level: Read-Only. Sync method: Periodic Polling.

Permissions (scopes) requested:

  • Mail.Read
  • User.Read
  • offline_access

Data we retrieve and index:

  • Inbox message bodies, threads, and conversation history
  • Sender details, timestamps, and attachment metadata

Data we never access:

  • Memora NEVER sends, alters, or deletes Outlook messages
  • Personal Microsoft accounts (only enterprise M365 supported)
  • Account recovery codes or sensitive credentials

Actions Memora can take: None. This connection is strictly read-only.

Discord

Connects community or developer Discord servers so teams can query technical FAQs and project lore.

Authentication: OAuth 2.0. Access level: Read + Confirmed Action. Sync method: Webhooks + Scheduled Sync.

Permissions (scopes) requested:

  • bot
  • applications.commands
  • messages.read
  • guilds

Data we retrieve and index:

  • Messages in assigned server channels
  • Thread conversations and user display names

Data we never access:

  • User private DMs outside bot conversations
  • Unshared server voice channels or server billing

Actions Memora can take (only after approval):

  • replyinthread (threaded answers with citations)

Element (Matrix)

Indexes encrypted Matrix rooms for privacy-conscious organizations seeking sovereign team collaboration.

Authentication: App Install. Access level: Read + Confirmed Action. Sync method: Periodic Polling.

Permissions (scopes) requested:

  • matrix:room_history
  • matrix:read_messages

Data we retrieve and index:

  • Decrypted room messages where end-to-end room keys are provided
  • Matrix event timestamps and member avatars

Data we never access:

  • Rooms where room keys have not been explicitly exported to Memora
  • User Matrix master recovery keys

Actions Memora can take (only after approval):

  • send_message (in-room responses)

WhatsApp & Telegram

Allows mobile team members to query workspace memory via chat apps on the go.

Authentication: API Token. Access level: Read + Confirmed Action. Sync method: Webhooks + Scheduled Sync.

Permissions (scopes) requested:

  • inbound_messages
  • send_message

Data we retrieve and index:

  • Direct inquiries sent to the Memora assistant contact number/bot
  • Conversational prompts and follow-up threads

Data we never access:

  • Personal contacts list or phone device storage
  • Calls or external chat history outside Memora conversations

Actions Memora can take (only after approval):

  • Direct message replies to user inquiries

2.2 Engineering

GitHub

Builds relational graphs between engineers, commits, PRs, and tickets to answer "who worked on what" and "why was this merged".

Authentication: OAuth 2.0. Access level: Read + Confirmed Action. Sync method: Webhooks + Scheduled Sync.

Permissions (scopes) requested:

  • repo:status
  • repo_deployment
  • public_repo
  • read:org
  • read:user
  • user:email

Data we retrieve and index:

  • Issue titles, bodies, labels, and resolution comments
  • Pull request discussions, reviews, and merge statuses
  • Commit messages, authors, dates, and diff snippets
  • Repository documentation (README, architectural specs)

Data we never access:

  • Repository write access to push commits or delete branches
  • Organization billing, SSO tokens, or private secrets/keys in repos
  • Unconnected private repositories outside selected spaces

Actions Memora can take (only after approval):

  • creategithubissues (strictly gated on explicit user command, requires approval)

Jira Cloud

Synchronizes project backlog and sprint progress into the workspace graph for real-time delivery status answers.

Authentication: OAuth 2.0. Access level: Read + Confirmed Action. Sync method: Webhooks + Scheduled Sync.

Permissions (scopes) requested:

  • read:jira-work
  • read:jira-user
  • write:jira-work
  • offline_access

Data we retrieve and index:

  • Jira issue summaries, descriptions, and custom fields
  • Sprint status, story points, and epics
  • Assignees, reporters, and activity comments

Data we never access:

  • Atlassian administrative access or billing profiles
  • Destructive deletion of tickets or projects

Actions Memora can take (only after approval):

  • createjiraissues (converts meeting recaps into drafted tickets; user review required)

GitLab

Enables engineering teams on GitLab to index repository history, issues, and code reviews into searchable facts.

Authentication: OAuth 2.0. Access level: Read-Only. Sync method: Webhooks + Scheduled Sync.

Permissions (scopes) requested:

  • read_api
  • read_repository
  • read_user

Data we retrieve and index:

  • GitLab issues, descriptions, and milestone timelines
  • Merge requests, review threads, and commit logs
  • Repository files in specified branch tracks

Data we never access:

  • CI/CD environment variables, secret tokens, or pipeline write access
  • Self-hosted runner configurations or administrative rights

Actions Memora can take: None. This connection is strictly read-only.

Trello

Surfaces task card statuses, checklists, and project roadmap milestones across agile marketing or sprint boards.

Authentication: OAuth 2.0. Access level: Read-Only. Sync method: Periodic Polling.

Permissions (scopes) requested:

  • read
  • account

Data we retrieve and index:

  • Board names, lists, cards, and checklists
  • Card descriptions, due dates, member assignments, and comments

Data we never access:

  • Power-Up administrative data or Trello Gold/Enterprise payment cards
  • Boards outside the authorized workspace selection

Actions Memora can take: None. This connection is strictly read-only.

2.3 Productivity

Google Drive (Docs, Sheets, Slides)

Extracts, chunks, and embeds company proposals, research documents, and design specs into semantic vector memory.

Authentication: OAuth 2.0. Access level: Read-Only. Sync method: Periodic Polling.

Permissions (scopes) requested:

  • https://www.googleapis.com/auth/drive.readonly
  • https://www.googleapis.com/auth/drive.metadata.readonly

Data we retrieve and index:

  • Document textual content (Google Docs, PDF text, presentation slides)
  • File names, folders, owners, sharing permissions, and timestamps
  • Spreadsheet cell contents and tabular summaries

Data we never access:

  • Drive files outside folders explicitly shared with or accessible to the user
  • File modification, overwriting, or deletion
  • Google Photos, videos, or raw binary downloads beyond text indexing

Actions Memora can take: None. This connection is strictly read-only.

Notion

Synchronizes your internal company wiki and project databases, allowing the agent to cite company policies and record notes.

Authentication: OAuth 2.0. Access level: Read + Confirmed Action. Sync method: Webhooks + Scheduled Sync.

Permissions (scopes) requested:

  • read_content
  • update_content
  • insert_content

Data we retrieve and index:

  • Page titles, blocks, headings, callouts, and paragraphs
  • Database records, schema properties, and linked relational entries
  • Page comments mentioning the Memora integration

Data we never access:

  • Pages and workspaces not explicitly shared with the Memora integration
  • Notion workspace settings, workspace members management, or billing

Actions Memora can take (only after approval):

  • createnotionpage (creates drafted recap pages or notes in designated databases)
  • add_comment (replies to comment mentions with cited knowledge)

OneDrive & SharePoint

Ensures Microsoft 365 enterprise documentation and SharePoint portals are continuously indexed into secure semantic memory.

Authentication: OAuth 2.0. Access level: Read-Only. Sync method: Periodic Polling.

Permissions (scopes) requested:

  • Files.Read.All
  • Sites.Read.All
  • offline_access

Data we retrieve and index:

  • Office documents (Word, Excel, PowerPoint) and PDF text
  • Document library structures, timestamps, and author identities

Data we never access:

  • Altering, moving, or deleting customer company files
  • Personal OneDrive files not shared with the corporate account

Actions Memora can take: None. This connection is strictly read-only.

2.4 Infrastructure

European Cloud Infrastructure (OVH, Scaleway, Hetzner)

All Memora server processes, vector embeddings, and database indexes operate on dedicated European bare-metal infrastructure under GDPR.

Authentication: App Install. Access level: Read-Only. Sync method: Periodic Polling.

Permissions (scopes) requested:

  • Bare Metal Colocation
  • Encrypted Persistent Volume
  • Private Subnet

Data we retrieve and index:

  • Internal container hosting and persistent ArangoDB multi-model storage
  • Encrypted backup volumes located exclusively within the European Union (France & Germany)

Data we never access:

  • No third-party cloud analytics or foreign intelligence access
  • No cross-border transfers outside the European Economic Area

Actions Memora can take (only after approval):

  • Internal storage only

3. How data ingestion and token storage work

  • Encrypted tokens. Access and refresh tokens are encrypted at rest with AES-256. They are never sent to browsers or returned in API responses.
  • Permission-aware answers. Memora only retrieves and cites content the asking user can already access in the source tool.
  • Disconnect and delete. Disconnecting a service immediately revokes its token. An administrator can delete all memories and vectors ingested from that source.

4. Frequently asked questions

Do you sell or share our data? No. We do not sell, license, or monetize workspace data. Ingested data is stored in tenant-isolated collections and used only to answer questions for your authenticated team members.

Can the agent delete our files or tickets? No. Destructive operations are not available to the agent. Write actions require confirmation on the chat platform.

Where are embeddings processed and stored? Embeddings are generated with self-hosted open models or zero-retention API tiers, and stored on our EU infrastructure in France and Germany.

5. Contact

For security questionnaires or questions about this page, contact security@celestical.eu.