Privacy Policy

1. Introduction

HostWebhook LLC, a Delaware limited liability company ("HostWebhook", "we", "us", "our"), operates hostwebhook.com. We provide workflow automation infrastructure: you receive events from webhooks, schedules, and connected apps, then route them through pipelines that transform, filter, enrich, and deliver the result wherever you choose. This Privacy Policy explains what we collect, why, who else touches it, and what you can do about it.

It covers the dashboard, the API, the CLI, the hosted chat and voice pages, and every node that runs inside a pipeline — and our public website (the home page, this policy and the terms), including what we learn about its visitors through Google Analytics if they allow it (section 10).

2. Two Kinds of Data

Almost every question about privacy on this platform has a different answer depending on which of these two you mean, so we separate them up front.

• Account Data — information about you as our customer: your email, your sessions, your billing record. For this data we are the controller and this policy describes what we do with it.
• Customer Data — everything that flows through your pipelines: webhook payloads, records pulled from connected apps, files, chat transcripts, and the configuration you build. Some of it may be personal data about people who are not our customers.

For Customer Data we act as a processor (service provider) and you are the controller. We process it only on your documented instructions — which, in practice, means the pipelines you build and run. We do not mine it, profile it, sell it, or repurpose it. You are responsible for having a lawful basis to send us that data and for telling the people it describes what you are doing with it.

3. Information We Collect

By category, with the purpose for each:

• Account and identity — email address, name, avatar URL, timezone, and (if you sign in with Google) your Google account identifier. Used for authentication, notifications, and support.
• Session and security — IP address, approximate country and city derived from it, browser User-Agent, client type (web, CLI, mobile), and last-used timestamp. Used to show you your active sessions, to let you revoke one, and to detect abuse.
• Pipeline configuration — nodes, transforms, filters, schedules, prompts, connections, and canvas layout.
• Event and run data — incoming payloads, the output of each node, delivery attempts, retries, and errors.
• Credentials — OAuth tokens, API keys, bot tokens, signing secrets, and database connection strings you save. Encrypted at rest and never returned to the browser once stored.
• Files — media you upload to the library or to a voice agent's knowledge base, attachments and media a trigger downloads, and files your pipeline nodes produce, plus derived thumbnails. How long each kind is kept is in section 8.
• Conversations — messages exchanged on a hosted chat page, transcripts of voice agent calls, and your conversations with the in-app assistant (Hermes AI and Hermes Builder).
• AI content — prompts, model responses, tool calls, token counts, and cost. See section 4.
• Usage and billing — event counts, AI credits consumed, plan tier, and your Stripe customer and subscription identifiers.
• Audit logs — who in your organization did what, and when a credential was read and by which node.
• Website analytics — only if a visitor allows it, and only on our public pages: the pages visited, a random browser identifier kept in the _ga cookies, approximate location, device and browser, how the visitor arrived, and interactions such as scrolling and time on the page. Collected by Google Analytics; see section 10.

We do not collect advertising identifiers, cross-site tracking data, biometric data, or precise geolocation. Google Analytics runs with its advertising features and Google signals turned off.

4. AI Features

HostWebhook uses AI in two distinct ways, and the difference matters because it changes whose account the data is sent under and who chose the provider.

4.1 AI we run for you (platform AI)

Some features in the dashboard call a model on our own account and bill you in AI credits. Today these are: rewriting a prompt you are drafting, building a pipeline from a description on the canvas, the in-app assistant chat, adapting and scoring social media captions per platform, and generating descriptions for MCP tools. These run on Anthropic's Claude models. What we send is the text you typed plus the surrounding context the feature needs — for the canvas builder, that includes the shape of your pipeline.

Your organization's owner, and anyone they allow, can review the conversations and memories of the in-app assistant (Hermes AI and Hermes Builder) for security. Your organization can decide whether you can see your own past Hermes AI conversations.

If you rate an answer from Hermes AI, our in-app assistant, we keep your rating and any comment you write. If you also tick "Include this answer", a member of the HostWebhook team may read that one answer — which can contain data from your connected accounts, including Google — only to improve Hermes. It is never used to train AI models, never shared with AI providers, and is deleted with your account.

4.2 AI you run yourself (the AI Node)

When you put an AI Node in a pipeline, it runs under your own credential with the provider you pick: Anthropic, OpenAI, Google, Groq, OpenRouter, or a self-hosted Ollama endpoint. Whatever the node receives — which can include the full event payload — is sent to that provider under their terms and privacy policy, not ours. Choosing a provider is choosing where that data goes. A self-hosted Ollama endpoint keeps it on infrastructure you control.

When you use OpenRouter, every request that carries Google user data asks OpenRouter to route only to providers that do not store your data or train on it (data_collection: "deny"). If a model has no such provider, the request fails instead of falling back.

4.3 Supporting AI services

• Embeddings — the memory of AI Nodes and of the in-app assistant sends text to OpenAI's embeddings API under our own paid OpenAI account, which does not let OpenAI train on it. Vector store nodes send text to OpenAI or Cohere with your credential.
• Voice agents — run on ElevenLabs, Vapi or Retell AI with your own account, plus OpenAI or ElevenLabs for speech synthesis. Call audio and transcripts are stored in object storage.
• Memory backends — conversation memory can live in our built-in vector store or in a backend you point at: your own MongoDB or your own PostgreSQL. You choose, with your credential.
• MCP servers — an AI Node can call tools on external MCP servers you connect. This is off by default and enabled per node; the node sends those tools only the arguments the model produces.

4.4 AI trace logging

So you can debug what a model actually did, we record each AI call: the prompt and system message, the response content, tool calls, token counts, latency, cost, and any error. These traces power the observability panel in your dashboard, are visible only to your organization, and are deleted on your plan's retention schedule along with the rest of your run history. To show you the quality of your AI responses, each trace is also scored automatically by Anthropic's Claude API under our own paid account. Anthropic's commercial terms do not allow it to train on that content.

4.5 What we do not do with AI

• We do not use your data — prompts, payloads, files, or configuration — to train, fine-tune, or evaluate any model, ours or anyone else's.
• The AI providers we use on our own account (Anthropic and OpenAI) do not train on what we send them under their API terms, and we keep their optional data-sharing programs disabled. For providers you connect with your own account, their terms and your plan apply: some free tiers (for example the unpaid Gemini API, or ElevenLabs outside its Enterprise plan) may use data to improve their models.
• We do not use AI to make automated decisions that produce legal or similarly significant effects about you. Every AI Node runs inside a pipeline you built, on data you routed to it, toward an outcome you defined.
• We do not profile you or your end users for advertising.

Model output can be wrong. Treat anything a pipeline generates as a draft that needs a human check before it reaches a decision that matters.

5. How We Use Your Information

We use what we collect to:

• Run your pipelines — receive events, execute nodes, and deliver results to the destinations you configured
• Authenticate you, keep sessions valid, and let you review and revoke them
• Send failure alerts, approval requests, system notices, and account email
• Meter usage against your plan and process billing
• Keep the service up: monitor health, diagnose failures, prevent abuse, and enforce rate limits
• Provide support when you ask for it
• Comply with legal obligations
• Understand how people find and use our public website, only for visitors who allow it

Where the GDPR applies, our legal bases are performance of a contract (running the service you signed up for), legitimate interests (security, abuse prevention, and service improvement), legal obligation (tax and accounting records), and consent for website analytics — which you can withdraw at any time (section 10).

6. Google API Data

When you connect a Google account via OAuth2 to use Sheets, Drive, Docs, Calendar, Gmail, Contacts, Tasks, Slides, or Analytics:

• We access Google data only to perform the operations configured in your pipelines, and only within the scopes you granted
• We store OAuth refresh tokens encrypted with AES-256-GCM
• We do not transfer Google data to third parties for our own purposes, use it for advertising, sell it, or allow humans to read it except with your explicit consent, for security purposes, to comply with applicable law, or in aggregated and anonymized form for internal operations
• We do not use data obtained through Google Workspace APIs — Gmail, Drive, Docs, Sheets, Calendar, Tasks, Slides, or any other — to develop, improve, or train generalized (non-personalized) AI and/or ML models
• HostWebhook's use and transfer of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements

Gmail

If you connect Gmail, HostWebhook asks Google for one Gmail permission, gmail.modify, plus your email address. We use it only for what your pipelines do with that mailbox:

• The Gmail trigger watches the mailbox and reads new messages — content, headers, labels, and attachments if you ask for them — to start your pipeline.
• Gmail actions search and read messages and threads, send email, replies, and drafts, add or remove labels, mark messages read or unread, and move messages or threads to Trash or back.
• AI agent tools, if you turn them on, let your own AI Node or MCP client do the same on your mailbox. You can block any tool.

We never delete mail permanently: "delete" moves it to Trash, and we do not request full mailbox access.

How long Gmail data is kept:

• Message data a pipeline reads stays in your run history for your plan's retention period (section 8), and is then deleted automatically.
• Attachments the trigger downloads are saved as temporary files, scanned for malware inside Cloudflare, and deleted automatically 24 hours after download.
• Beyond that, Gmail data is stored or sent elsewhere only where your own pipeline puts it.
• Deleting the credential revokes HostWebhook's access with Google. Deleting your account deletes everything it owns (section 8).

Who handles Gmail data: only the infrastructure providers that run the platform, listed in section 9 — Railway (compute), MongoDB (run history), Redis (job queues), Amazon Web Services (the in-app assistant and AI traces, only if the assistant or an AI Node in your pipeline receives it), and Cloudflare R2 (attachments) — which process it only to run the service. We do not share it with anyone else ourselves; the only other places it goes are the ones your pipeline sends it to, as explained next.

One thing worth stating plainly: if you build a pipeline that takes data out of Google and sends it somewhere else — into an AI Node, an HTTP Action, a database, a Slack channel — that transfer happens because you configured it. You direct it, you are responsible for it, and it is not a transfer we make for our own purposes. You can see every such route on your canvas, and remove it there.

Google data and AI

Google user data is sent to an AI model only when your own pipeline does so — for example an AI Node that reads a Google Sheet as a tool, or a sheet's output routed into an AI Node — and only to services that do not use it to train or improve their models. In particular: with the Gemini API, an AI Node receives Google user data only if you confirm on its Google AI credential that the key is from a Google Cloud project with billing enabled (the free tier uses what it receives to improve Google's products); Google user data is never sent to voice agents or to Cohere embeddings; and every OpenRouter request that carries Google user data is restricted to providers that do not store or train on it. If we cannot tell whether a step carries Google user data, we treat it as if it does. We do not self-host AI models. If you connect your own Ollama server, the data is processed only on your server and is never shared with a model provider for training or any other purpose.

HostWebhook staff never read your Google user data, except a single Hermes answer you explicitly choose to send with your feedback.

HostWebhook's use and transfer to any other app of information received from Google Workspace APIs adheres to the Google Workspace API User Data and Developer Policy, including its Limited Use requirements.

7. Data Storage & Security

• Records are stored in MongoDB with encryption at rest; files live in S3-compatible object storage
• Credentials — API keys, OAuth tokens, signing secrets, connection strings — are encrypted with AES-256-GCM and are never sent back to the browser after you save them
• All connections use HTTPS/TLS in transit
• Outbound requests to addresses you supply are validated against SSRF attacks, and databases that are not publicly reachable are contacted over a tunnel rather than by opening them to the internet
• Access to your data is scoped to your organization, and every credential read by a node is logged
• Account passwords are hashed; we never store them in a form we can read

No system is perfectly secure. We follow industry practice for access control and infrastructure security, and we keep the blast radius of any single failure as small as we can.

8. Data Retention

Operational data is deleted automatically on the schedule below. Retention scales with your plan:

• Audit and access logs are kept for at least 30 days on every plan, including Free. A plan can raise that floor but not lower it — these logs are how a security question gets answered after the fact.
• Credentials and OAuth tokens are kept until you delete them or revoke access at the provider
• Files in your media library, and files attached to a voice agent's knowledge base, stay until you delete them or delete the workspace holding them
• Every other file is temporary and is deleted automatically 24 hours after it is created — for example, the attachments and media a trigger downloads (Gmail, Telegram, WhatsApp, OneDrive, Word) and the files your pipeline nodes produce
• Account and organization records are kept for as long as the account exists
• An organization that is frozen (a payment kept failing after a 14-day grace period, or its paid plan ended while you already own a free organization) becomes read-only and its data is kept for 30 days, with email reminders, and then deleted for good unless you pay to reactivate it
• Invoices and billing records are kept as long as tax law requires, even after deletion
• Website analytics are kept by Google Analytics for at most 14 months; its _ga cookies expire after 13 months, and your answer to the analytics notice is asked again after 12

Deleting your account is immediate and permanent. There is no soft delete and no undo: your sessions and your user record are destroyed before the request even returns, and the cascade that removes everything you own — organizations, workspaces, pipelines, credentials, run history, and stored files — runs right behind it, and every subscription you pay for is cancelled. Deleting a single organization works the same way, from its owner, and it also cancels that organization's subscription. Backups age out on their own cycle within 30 days.

9. Subprocessors

These providers operate on every account, because they are how the platform runs:

These receive data only if you connect them. Adding a credential or a node is what starts the flow; removing it stops it:

• OpenAI — AI Node completions, vector store embeddings, speech synthesis, and audio transcription with your own API key
• Google — Sheets, Drive, Docs, Calendar, Gmail, Contacts, Tasks, Slides, Analytics, and Gemini models
• ElevenLabs — Speech synthesis for voice agents
• Cohere — Embeddings for vector store nodes with your own API key
• Vapi, Retell AI — Voice agents with your own account
• Slack, Discord, Telegram, WhatsApp — Message delivery and trigger events
• X, LinkedIn, Bluesky, Mastodon, Threads, Instagram, Facebook — Publishing and reading social media posts
• Notion, Atlassian (Jira, Confluence, Trello), ClickUp, Asana — Reading and writing records in your workspace
• Firecrawl — Web page scraping and extraction
• AWS S3 — Object storage in a bucket you own
• Anthropic, Groq, OpenRouter — AI Node completions with your own API key
• Microsoft (Outlook, Outlook Calendar, OneDrive, Excel, Word, Teams) — Reading and writing in the Microsoft 365 account you connect
• Stripe (the account you connect) — Reading and writing customers, payments, subscriptions, invoices, products and prices, and receiving its events
• Shopify, WooCommerce, Calendly, Airtable, Mailchimp — Reading and writing records in the accounts you connect, and receiving their events
• GitHub — Reading and writing issues, pull requests, and workflows in the repositories you connect
• Apify — Running actors and reading their results with your token
• Databases you connect (PostgreSQL, MongoDB) — Running the queries and writes your pipeline defines
• MCP servers you connect — Tools an AI Node may call, scoped per node

Google Analytics (Google LLC) receives data only from visitors of our public pages who allow analytics — never account or pipeline data (see section 10).

Each provider has its own privacy policy and its own terms. We send each one the minimum the integration needs to work, and nothing else. We will update this list before adding a subprocessor that touches personal data.

10. Cookies and analytics

When you are signed in, we use two cookies, both strictly necessary, both httpOnly so no script on the page can read them:

• accessToken — proves who you are on each request. Short-lived.
• refreshToken — issues a new access token so you stay signed in. Expires after 7 days of disuse.

On our public pages — the home page, this policy and the terms — we also use Google Analytics, and only if you allow it. A notice asks first; until you answer, and if you say no, nothing from Google loads and no analytics cookie is set. If you say yes, Google Analytics sets its _ga cookies (a random identifier for your browser, kept for 13 months) and records which of those pages you visit, how you arrived, roughly where from (Google uses your IP address to work out country and city, and does not store it), your device and browser, and interactions such as scrolling and time on the page — so we can see how people find the site. Its advertising features and Google signals are turned off, and it never runs inside the dashboard, the hosted chat or the widgets our customers embed. Your answer is kept in your browser's local storage, not in a cookie, and we ask again after 12 months.

If your browser sends Global Privacy Control, we treat it as a no: the notice does not ask and analytics stays off. You can change your answer at any time, and saying no later also deletes the _ga cookies: open cookie settings.

We run no advertising pixels and no session recording. Clearing the two cookies above simply signs you out.

On our public site, our hosting and security provider, Cloudflare, also sets a cf_clearance cookie once it has checked that your browser is not an automated bot, so you are not checked again on every visit. It is strictly necessary for the site's security, so it does not need your consent. It is httpOnly, is not used for analytics or advertising, and expires after one year.

11. Your Rights

Wherever you are, you can:

• Access your personal data and pipeline configuration
• Correct anything inaccurate, from your profile settings
• Delete your account and everything it owns
• Export your data in a machine-readable format
• Revoke any OAuth connection or session at any time
• Object to or restrict processing based on legitimate interests
• Complain to your local data protection authority

California residents (CCPA/CPRA)

You have the right to know what we collect and why, to delete it, to correct it, to opt out of sale or sharing, to limit the use of sensitive personal information, and not to be discriminated against for exercising any of them. Categories collected and their retention are listed in sections 3 and 8.

We do not sell your personal information, and we do not share it for cross-context behavioral advertising — not in the last 12 months, and not as a matter of how this product is built. There is no advertising business here to sell it to. We honor Global Privacy Control signals: with one on, the analytics on our public pages stays off (section 10).

To exercise any of these rights, use the dashboard settings or write to support@hostwebhook.com. We answer within 30 days. If you are asking about data inside another company's pipeline, we will point you to that company — for Customer Data they are the controller and we act on their instructions.

12. International Transfers

Our infrastructure runs in the United States. HostWebhook LLC is organized in Delaware, United States, and the service is operated from Mexico. If you are in the European Economic Area, the United Kingdom, or Switzerland, using the service means your data is transferred to and processed in the United States. Where required, we rely on the European Commission's Standard Contractual Clauses for those transfers. Any node you configure to call a service in another region will send data there as well.

13. Children's Privacy

HostWebhook is a developer tool and is not directed at children. We do not knowingly collect personal information from anyone under 16. If you believe a child has given us personal information, write to us and we will delete it.

14. Security Incidents

If a breach affects your personal data, we will notify you without undue delay and, where the law sets a deadline, within it — 72 hours of becoming aware, under the GDPR. The notice will say what happened, what data was involved, what we have done, and what we recommend you do. If you believe you have found a vulnerability, report it to support@hostwebhook.com and we will respond.

15. Changes to This Policy

We may update this Privacy Policy from time to time. We will notify you of any material changes by posting the new policy on this page and updating the "Last updated" date, and for changes that meaningfully expand what we collect or who receives it, by email. Your continued use of the service after changes constitutes acceptance of the updated policy.

16. Contact Us

If you have any questions about this Privacy Policy or our data practices, contact us at: support@hostwebhook.com

HostWebhook LLC
8 The Green, Ste A
Dover, DE 19901, United States