Publive MCP Connector — Privacy Policy
Last updated: 9 September 2026
Applies to: the Publive MCP connector (mcp.thepublive.com) — the integration that
lets an AI assistant such as ChatGPT or Claude read and manage your Publive content on
your instruction.
1. Who we are
Publive ("we") operates the MCP connector described here. The connector is a bridge: it holds your Publive API credentials and uses them to call the Publive CDS and CMS APIs on your behalf when you ask an AI assistant to do something with your content.
This policy covers the connector specifically. It is narrower than, and does not replace, Publive's main product privacy policy — for anything about the Publive platform itself, see that policy.
2. What we collect
We collect only what is needed to authenticate you against your own Publive account.
| Data | Why | Source |
|---|---|---|
| Publisher ID | Identifies which Publive tenant to address | You, at sign-in |
| CDS API key + secret | Authenticates read calls to the Publive CDS API | You, at sign-in |
| CMS API key + secret | Authenticates read and write calls to the Publive CMS API | You, at sign-in |
OAuth client_id and redirect URI |
Identifies the AI client you connected | Registered by the client |
| Access and refresh tokens | Keeps you signed in without re-entering credentials | Generated by us |
| Operational logs | Diagnosing failures, detecting abuse | Generated by us |
Either credential pair may be omitted. A CDS-only sign-in produces a connection that can only read; the tool surface narrows to match.
Reader data
The connector holds no tool that handles a reader's personal data. There is no way, through this connector, to create or change a reader account, read or alter anyone's newsletter subscription, or trigger a password-reset or email-verification message. The tools that could do those things were removed rather than gated: their effect was an email already sent to a third party by the time the call returned, so no confirmation step could have made them safe to drive from a conversation.
Publishers manage reader accounts and subscriptions in the Publive dashboard. If a reader needs to sign up or reset a password, direct them to your own site's flows.
fetch_newsletter_groups reads your own newsletter group configuration — names,
logos, descriptions — and returns no subscriber and no email address.
3. What we do NOT collect
We do not collect, store, or process your conversations with the AI assistant. The
connector strips the prompt and _prompt arguments from every tool call before the
tool runs, so prompt text is discarded at the entry point rather than logged or
forwarded.
We also do not collect: your name, your AI provider account, billing information, or any content beyond what a tool you invoked explicitly requested from your own Publive instance.
4. How your credentials are stored
Your Publive API keys and secrets are encrypted at rest with Fernet (AES-128-CBC with HMAC-SHA256 authentication) before being written to the database, using a key held in AWS Secrets Manager and never stored alongside the data it protects. The encryption supports key rotation without downtime.
All storage is in AWS ap-south-1 (Mumbai, India). See §9.
Credentials are transmitted only over TLS, to the Publive CDS and CMS APIs, using HTTP Basic authentication — the same way you would call those APIs yourself. Each service is sent only its own key pair; the two are never interchanged.
5. What the connection can do
The connector exposes 61 tools: 21 read-only CDS tools and 40 CMS tools. What any given connection can reach is bounded twice over:
- By credentials — a connection holds only the key pairs you supplied, and the tool surface narrows to match. Supplying only the CDS pair produces a genuinely read-only connection: the 40 CMS tools are neither offered nor callable, because there is no CMS credential for them to spend.
- By confirmation — see below.
Content-changing operations require an explicit confirmation step, and deletions require two (a preview, then an explicit delete confirmation). Nothing is created, changed, or deleted without a deliberate second call.
6. Who we share it with
We do not sell your data and we do not share it for advertising.
| Recipient | What they receive | Role |
|---|---|---|
| Amazon Web Services | All hosted data (ap-south-1) | Infrastructure processor |
| New Relic | Operational telemetry: tool names, response times, error categories, publisher IDs. No credentials, no tokens, no content. | Monitoring processor |
| Your AI provider (OpenAI, Anthropic, …) | The tool results you asked for | The client you chose to connect |
Your AI provider's handling of the results is governed by their own privacy policy, not this one.
7. Retention
| Data | Retained |
|---|---|
| Authorization codes | Single-use, and unusable after 10 minutes |
| Access tokens | Issued with a 1-hour lifetime; replaced by refresh |
| Refresh tokens | Until you revoke or disconnect |
| Browser sessions | Until you sign out, or the session's absolute lifetime elapses |
| Operational logs | 30 days, then deleted |
An authorization code is spent the moment it is exchanged and cannot be exchanged twice; after ten minutes it is refused regardless. Revoking a connection deletes its record outright rather than merely invalidating the token — see §8.
You can have every record we hold for you deleted at any time by revoking, or on request (§11).
8. Deleting your data and revoking access
You can end a connection at any time, and doing so deletes the stored credentials:
- From your AI assistant — disconnect the Publive connector.
- Programmatically —
POST https://mcp.thepublive.com/revokewith the token (RFC 7009). This deletes the record, not just the token. - In a browser — sign out at
https://mcp.thepublive.com/connect. - By request — contact us (§11) and we will delete any remaining records.
Revoking does not delete content in your Publive CMS. Content changes made through the connector are changes to your own Publive instance and are managed there.
9. Where your data is processed
All processing and storage happens in AWS ap-south-1 (Mumbai, India). There is no EU or US region.
The connector is offered to publishers established outside the EU and EEA. It is not made available to publishers established in the EU or EEA, and it is not intended to process the personal data of individuals in the EU or EEA — so no Chapter V transfer mechanism is engaged.
If you are established in the EU or EEA, or you need reader data processed inside the EU or EEA, contact us (§11) before connecting rather than signing in.
10. Related documents
The terms of service govern your use of the connector; the connector documentation describes every tool and what it can reach.
11. Contact
Questions, data requests, or security reports: https://www.thepublive.com/contact-us
12. Changes
We will update the "Last updated" date when this policy changes. Material changes affecting how credentials are stored or shared will be announced before taking effect.