Privacy Policy

Last updated: October 25, 2026

This policy describes what Podclave LLC, a Florida limited liability company (“Podclave,” “we”), collects when you use podclave.com, why, and what we do with it.

We keep account information, connected credentials, and configuration to run the service. We also store Workspace files, setup, and agent history so your work can return between Sessions, plus the current content of World Shared files and My files. Private work is restricted to its owner through Podclave; shared content follows the access rules of the feature you use.

When you ask an agent to use connected Google Drive or Slack content, that content passes through Podclave. Requested Google Doc text and Slack message text, file details, and available previews or transcripts are screened by Google Cloud before they reach the agent. Requested Slack files are downloaded into your private Workspace for your agent to process; the original file contents are not screened by Google Cloud. Results can then reach your selected agent provider and remain in your Workspace’s history or saved files. We do not independently archive the source channels or documents. Content included in a saved conversation follows that conversation’s retention rules.

Our product analytics use no cookies. We do not sell your personal data. Optional voice dictation sends mic audio to ElevenLabs, which can retain audio and transcripts. Nothing is sent to ElevenLabs if you do not use the mic.

Questions or requests: hello@podclave.com.

1. What we collect and store

Account data. Your email address, your role, Lab and World memberships and invitations, email delivery status, and timestamps of account activity. Sign-in uses one-time codes sent to your email. If you register a passkey, we store the information needed to verify it. We do not receive or store your biometric data; that check happens on your device.

Credentials you connect. We store the agent and service credentials you choose to connect, including Anthropic (Claude), OpenAI (Codex) sign-in or API access, SpaceXAI (Grok), and personal GitHub access. We use them to authenticate the sessions you launch, refresh access where the provider supports it, and show your provider subscription usage.

We keep these credentials encrypted while stored by Podclave and exclude them from product analytics and application logs. Credentials needed by a tool can be made available to that tool in your Workspace. You can disconnect each provider in the product and revoke its authorization at the provider independently of us.

World repository access. A World can connect a GitHub App installation or Bitbucket repository access tokens. A World Admin can also give all World members read-only access to selected repositories with the Admin’s own GitHub sign-in or Bitbucket API token; that credential stays outside members’ Workspaces. We store the installation, these tokens encrypted, and selected repository information, and use the access granted to prepare repositories and support requested Git operations. This is separate from each member’s personal agent credentials.

Google Drive connection credentials. A World Google Drive connection works differently. Podclave creates a dedicated Google service account for that connection. A World Admin adds its email address to one Shared Drive and chooses its Google role. We keep the connection credential encrypted while stored and use it only for the Drive requests described in this policy. We do not send it to a Workspace, agent, or model provider, or include it in analytics or logs.

Configuration you store. World settings, selected repositories, tools, shared environment values (stored encrypted and supplied to every member’s Workspace), member grants, Workspace and Session names and state, and Public URL settings. For connected services, we store the connection settings, access level, and state. We also keep limited operation records for troubleshooting. These records do not contain a separate copy of Drive files or Slack message text.

Workspace work. We store your repository and other Workspace files, configuration, and agent history with our compute provider, including saved state needed to return after compute parks or is replaced. Other World members and World Admins cannot inspect your personal Workspace or its chat titles and contents through Podclave. Content you place in Shared files, send to a connected service, or expose through a Public URL follows that feature’s access rules.

Saved conversations. We keep a readable copy of your World conversations and retained attachments so you can review them while compute is parked. This copy excludes thinking panels and raw tool output. Other members and World Admins cannot read it through Podclave.

World Shared files and My files. If you use these folders in a World, Cloudflare R2 stores their current content and paths. Shared files are available to all members of that World. My files are private to one person in one World; other World members, including World Admins, cannot open them through Podclave. You can use these files from your Workspace, from browser file pages, or through an Agent access connection that you approve, without a running Session. A connection acts only as the identity it was approved for and has that identity’s current World access: it can read and write the World’s Shared files and that identity’s own My files, never another person’s My files. The client you connect receives the content it reads. Podclave does not keep prior file versions. An overwrite or delete can remove the prior content.

Agent access connections. An outside agent that you connect and approve can also start and manage your Sessions in the approved Worlds, read their messages, send messages and files, and receive Session event notices at the address its app provides. The connected app receives the content it reads under its own terms. You can revoke a connection at any time.

Billing data. Stripe handles payments and acts as merchant of record for Worlds plans and AI funds bought in Podclave. Card numbers go directly to Stripe and never touch our servers. We store the billing and usage records needed to manage your service.

AI usage reports. When the Lab pays for agents, we collect provider-reported AI costs, confirmed AI dollar charges, and any earlier credit charges. If a Lab Admin turns on automatic AI purchases, we keep the selected amounts, the Admin who accepted the charge terms, and the time of acceptance. Lab Admins can see the Lab’s billing totals and AI purchase and refund history. Agent usage reports can also include member email, model, effort setting, token counts, funding source, and cost estimates. Members see their own use. Lab Admins can see the Lab’s use, and World Admins can see use within their permitted World and paying Lab. These reports do not include prompts, responses, thinking, or tool output, and do not give admins access to private chats.

Operational logs. Standard server logs for security and debugging, which can include IP addresses and request paths. We don’t build profiles from them.

2. How content moves through the service

When you browse, upload, or download files, use a terminal, or chat with an agent, we and our hosting providers process the content needed to deliver that feature. Workspace files and agent history are stored as described in section 1. Files you upload or attach are placed in your Workspace. Temporary service copies used for a transfer are deleted after the transfer or a short expiry period. We do not keep a separate service archive of terminal output or file listings.

Because we supply the hosting, authorized service operators can technically access stored work. We limit access to what is needed to operate the features you request, investigate security or abuse, meet legal duties, or provide support with your permission. Private Workspace access restrictions apply to other customers and World Admins; they do not mean the hosting provider cannot process the data.

Agent providers. Prompts, responses, and content supplied to Claude, Codex, or Grok are processed by Anthropic, OpenAI, or SpaceXAI under the account access you select. Podclave and its hosting providers handle the credentials and traffic needed to deliver those sessions. The agent provider’s own retention and data-use settings apply to its copy. Agent history can also remain in your Workspace. Downloaded conversation files contain private content and attachments; anyone you give the file to can read those copies.

Lab-funded AI. If a World Admin permits this option and you select Lab funding for a Session, requests pass through Vercel AI Gateway and the providers that serve the selected model. This includes prompts, responses, and files or tool results the agent includes in a request. This access uses Podclave’s provider account, so the settings of your personal subscription do not control these requests. Provider retention, data-use rules, and processing locations apply to their copies. Choosing a subscription does not automatically authorize Lab-funded access. When a World Admin permits Lab funding, an agent in any of your Sessions in that World can also send text it selects to a decision model through Vercel AI Gateway, paid from the Lab’s AI funds.

Connected Google Drive data passes through Podclave and Google Cloud. When a World member asks an agent to use a connected Shared Drive, Podclave sends the requested search, read, create, or edit operation to the Google Workspace APIs. File and Drive ids, names, metadata, search terms, requested Google Doc text, and content supplied for a create or edit can pass through our service to complete that operation.

For a Google Doc read, we send the requested text to Google Cloud Model Armor in the United States to check for prompt injection before we return it to the agent. We send only the text, not your identity or account details. We do not store Drive search results or file content in our service records or application logs.

The screened text and other tool results are returned to the requesting agent. They can be sent to that member’s selected agent provider and can remain in the Session history. When a member asks an agent to create or edit Drive content, Google stores the resulting file or change in the connected Shared Drive. Every World member can use the tools that the World Admin enabled by the Google role.

The selected agent provider’s own retention and data-use settings apply to the copy it receives. A member must use Drive tools only with a provider plan that excludes general model training or after the member turns training off for that provider. The member must not send provider feedback about a conversation that contains Drive data if that feedback can be used for model training.

Connected Slack discussion passes through Podclave. A World member can ask an agent to list the connected bot’s public channels and read their messages and threads and retrieve attached files when file access is enabled. We retrieve the requested channel information, message text, author references, timestamps, source links, file details, and available previews or transcripts under that bot’s access. We send this text and these details to Google Cloud Model Armor in the United States for screening before returning them to the agent. Requested original files, including images, audio, video, and documents, pass through Podclave into the requesting member’s private Workspace; Podclave does not keep a copy. Original file contents are not sent to Google Cloud for screening. Your agent can process them with its available tools. You can delete downloaded files from your Workspace; otherwise, they follow your Workspace’s retention rules. The results can reach the member’s selected agent provider and remain in Session transcripts or saved work under that provider’s retention and data-use settings. We do not keep a separate message archive or log message text in our service records.

Every World member can use this access, even if that person is not a member of the Slack workspace. Add the bot only to public channels intended for that World. When a World Admin enables posting, members can ask agents to send messages and thread replies as that bot. We send the requested text to Slack, where channel members and connected apps can access it. Posts can include formatting, links, and mentions that notify the selected recipients. When user lookup is enabled, members can ask agents to find Slack workspace members by name. We retrieve their names and Slack identities to select the correct recipient; we do not request email access.

For each connected bot, we keep a list of the Slack workspace members whose messages that bot has read for the World: their Slack identity, display name, first and last time seen, and whether a World Admin has allowed their messages to direct work in the World. We also keep references to messages the bot read: their channel, time, and author, without the message text. We use this to decide whose messages can direct agent work. World members can see the list; World Admins change its permissions. Disconnecting the bot or deleting the World deletes the list and the references. Slack controls those messages under the workspace’s agreement and settings. The Slack connection does not read private channels or direct messages, search all Slack messages, or reply automatically to incoming messages.

We store the bot name and any icon you choose or upload so World members can identify the connection. Saved icons remain with the connection and are deleted when you remove an unfinished setup or delete the World. If you upload the icon to Slack, Slack controls that copy under its own settings.

Optional voice dictation goes directly to ElevenLabs. The first time you tap the mic in an agent chat, Podclave shows a separate consent notice. If you choose Use voice dictation, your browser stores that choice locally. When you tap the mic, your browser sends mic audio directly to ElevenLabs and receives the transcript directly from ElevenLabs. Podclave does not keep a copy of the mic audio. If you send the transcript as a chat message, it becomes part of your agent conversation and is stored with that history.

ElevenLabs uses default retention for this feature. It can retain the mic audio and transcript to provide and improve its service, troubleshoot, moderate content, and keep its systems secure. Do not use voice dictation for passwords, API keys, health data, financial data, or other sensitive information. There is no local speech-recognition fallback: if you use the mic, this ElevenLabs processing applies. If you type instead and do not use the mic, it does not apply.

Browser notifications. If you turn on notifications, we store your browser’s push subscription and its user-agent string (browser and device type). Notification content, such as the chat name, a short part of an agent’s question, approval request, or error, and the account it is for, is encrypted for your device and passes through the push service your browser uses (for example Apple, Google, or Mozilla).

3. Product analytics

We measure signups, product use, how long Sessions take to start, and homepage visits with event records that our servers keep in our own database in the United States. We do not send these events to an analytics provider. We do not use analytics cookies, session replay, or cross-site tracking.

  • Events are things like “user signed up”, “session opened”, and how long a Session took to become usable — business moments, identified by a pseudonymous account id, never your email.
  • Homepage visits, and when the homepage video starts and reaches its end, use a temporary visit identifier that is not linked to your account.
  • To measure Session start time across a page reload, your browser keeps a random attempt identifier in tab storage until the result is reported.
  • A link in some of our posts can carry a fixed tag that names the account that posted it. The tag passes through the sign-in forms so we can count verified signups from that link. We do not collect the full referring address or arbitrary link parameters. Untagged activity remains unattributed.
  • Events never contain: email addresses, code or file contents, file names or paths, repository names or URLs, credentials, shell input, agent prompt text, or IP addresses.

We use Sentry for operational monitoring. It receives error and performance data needed to diagnose service problems. We remove request parameters, headers, cookies, user details, and IP addresses before sending. Error messages and request paths can contain record ids, Lab or World identifiers, or other details of the failed operation. Sentry stores this data in the United States.

4. How we use data

To operate the service (everything in sections 1–2 exists to make a feature work, including transcribing mic audio when you choose voice dictation); to send transactional email (sign-in codes, a welcome message, invitations, and billing notices); to bill for service usage; to understand product usage in aggregate and fix errors; to secure the service and enforce our Terms; and to comply with law. We do not sell or rent personal data, and we do not use your content or your data to train machine-learning models.

Podclave’s use of information received from Google Workspace APIs adheres to the Google User Data Policy, including its Limited Use requirements. We use this information only to provide the Drive tools that a World member requests. We do not use raw or derived Google Workspace data for advertising, credit decisions, sale, or to create, train, or improve a general machine-learning or artificial-intelligence model.

5. Who we share data with

Only the vendors it takes to run the service, listed with their roles on the Subprocessors page — service hosting and compute (Fly.io, Vercel, and Hetzner), AI model routing for Lab-funded AI (Vercel AI Gateway), database (Neon), email (Resend), payments (Stripe), operational monitoring (Sentry), and bot protection, Public URL access and traffic, DNS, and World file storage (Cloudflare). Google Cloud supplies the dedicated Drive service accounts and screens requested Google Doc text and Slack message text, file details, and available previews or transcripts with Model Armor. ElevenLabs provides optional voice dictation only after the mic-specific consent described in section 2. Each receives only what its function needs.

Separately, when you connect Anthropic, OpenAI, SpaceXAI, GitHub, Bitbucket, Google Drive, or Slack, we interact with those providers under the authority you or your World Admin supplied — that is the feature. Your relationship with each provider is also governed by its privacy policy and terms. Requested Drive and Slack tool results can be sent to the agent provider selected for that session. If you turn on browser notifications, your browser’s push service carries their encrypted content, as described in section 2.

We will disclose data if validly compelled by law. Unless legally prohibited, we will tell you before handing over anything about you. If Podclave LLC is ever acquired or its assets sold, this policy continues to apply to data collected under it, and we will notify you before any change.

6. Where data lives

We are a US company. Our service data and hosted compute (listed on the Subprocessors page) are US-based. Vercel runs World Workspaces in the United States. World Shared files and My files use a US-jurisdiction Cloudflare R2 bucket, which stores and processes those objects in the United States. Retained conversation attachments and temporary conversation downloads use the same storage location. Cloudflare’s global network also handles bot protection, DNS, and traffic for apps exposed through Public URLs. Monitoring data sent to Sentry is stored in the US. Google Cloud Model Armor screens requested Google Doc text and Slack message text, file details, and available previews or transcripts in the United States. Your Google Workspace agreement and settings govern where Google stores the source Shared Drive. Slack and the agent providers process their copies under their own service terms and regional settings. ElevenLabs states that it has service locations in the United States, the Netherlands, and Singapore, and that personal data is transferred to the United States for storage. If you use Podclave from outside the United States, your service data is processed in the US and optional mic data can also be processed in those ElevenLabs locations. For users in the EEA/UK: we rely on our subprocessors’ standard contractual protections for applicable transfers, and the rights in section 8 are available to you regardless of where you live.

7. Retention and deletion

We keep account data, configuration, and credentials while needed to supply your service, subject to deletion requests and required billing or legal records. Disconnecting a personal credential stops our use of it. Its encrypted record remains in our database until your account is deleted. A World Admin’s Bitbucket token shared with a World stays encrypted in our database after it is revoked and after the World is removed. Revoke access at the provider as well if you need to invalidate a credential.

Workspace work. Parking compute preserves saved work. Running out of credits early does not start a storage-deletion grace period: saved Workspace work remains through the paid period. If that period ends without renewal, there is a seven-day storage grace period. After grace, we may delete stored Workspace work; continued retention is not guaranteed. This is not a promise of deletion at an exact time. Removing a Workspace permanently removes its saved work. Keep independent copies of anything you need.

AI usage history. Detailed AI reports cover the most recent 13 months. After that period, we remove model and usage detail for use that did not produce a customer charge. We keep the records needed to explain charges with billing records. Removing a Session or World does not erase those billing records. Usage numbers can also remain in your saved Workspace until it is removed.

Saved conversations. Parking compute does not delete saved conversations. Removing a Session removes its service conversation copy. Removing the World removes these copies for its Sessions. Conversation download links expire after 24 hours, and their temporary service copies are then removed. Files already downloaded or sent to an agent provider remain with those copies under their own retention rules.

World Shared files and My files. We keep their current content while the World retains those files. Deleting a file removes its current stored content; Podclave does not keep prior file versions. A Lab Admin can remove a World after its Workspaces are removed and its Drive connections are disconnected; removing it deletes its Shared files and every member’s My files for that World. These features are not a versioned backup service. The removed World’s configuration records, including encrypted access tokens and environment values, stay in our database; email us to have them deleted.

Disconnecting a Google Drive blocks new tool calls and deletes its dedicated Google service account and credential. The World Admin then removes that account from the Shared Drive and confirms this in Podclave; we keep the account’s email and connection state until then. Disconnecting does not delete or change files in Google Drive. Drive content already returned to an agent can remain in that Session’s history until the Session, its Workspace, or the World is removed.

For Slack, we store connection names, workspace and bot identities, access state, and encrypted app credentials and bot tokens. Disconnecting stops future reads, file downloads, and posts and deletes our bot token. We retain the app credentials and connection settings so a World Admin can reconnect through Slack. Deleting the World removes those records and credentials. Disconnecting does not uninstall the app from Slack or erase content already returned to a Session. Removing the bot from a Slack channel stops new reads, file downloads, and posts through that channel. Posts already sent remain in Slack until removed there. We keep send records with the requesting member, destination, time, and result. They do not contain the posted text. Deleting the World removes its send records.

To delete your account, email hello@podclave.com and we will delete your account and the service data above within 30 days, except minimal records we must keep (for example invoices, which Stripe retains for tax law). Your private Workspace work is included in that request. Retrieve anything you want to keep first. Content shared with a team or sent to another service can remain there under its access and retention rules; contact us to include that content in your request where you have the right to remove it.

ElevenLabs controls its own retention of voice-dictation audio and transcripts under its privacy policy. Email us if a privacy request includes voice dictation, and we will pass on or support the request as required.

8. Your rights

Email hello@podclave.com to access, correct, export, or delete your personal data, or to object to a use of it. We honor these requests for everyone, not only where a statute (GDPR, UK GDPR, CCPA) requires it, and we respond within 30 days. We will never discriminate against you for exercising them. We do not sell or “share” personal information as the CCPA defines those terms.

9. Security

We encrypt credentials and other sensitive values while stored and use encrypted connections when data moves over a network. Sign-in uses one-time email codes or passkeys. Production access is limited to people who operate the service. No system is perfectly secure; if we suffer a breach affecting your data, we will notify you promptly and tell you what we know.

10. Children

Podclave is not directed at children under 13, and we do not knowingly collect their data (where your country sets a higher minimum age for online services, that age applies instead). Users under 18 participate as invited members of a Lab managed by an adult — see the Terms, section 3. If you believe a child below the minimum age has an account, email us and we will delete it.

11. Changes

We will update this policy as the product evolves. The “Last updated” date reflects the current version, and prior versions are available on request. For material changes we will notify you by email or in the product before they take effect.

Podclave LLC · Florida, USA · hello@podclave.com

© 2026 Podclave LLC · Operated by Podclave LLC, a Florida limited liability company · hello@podclave.com