4Degrees
Sub-processors
This page lists the third-party services that process data on 4Degrees’ behalf when your firm uses the 4Degrees AI Connector and the 4Degrees Slack app. It supplements, but does not replace, the sub-processor and data-processing terms in our master 4Degrees Privacy Policy and Data Processing Agreement, which govern your firm’s overall use of the 4Degrees platform.
Effective April 27, 2026Scope #
The sub-processors below are those engaged for the connector and Slack-app surface specifically
— the Model Context Protocol (MCP) server and Slack integration hosted at
https://sa-mcp.4degrees.ai/mcp. Sub-processors for the broader 4Degrees platform (for
example payments, transactional email, or analytics) are governed by our master Privacy Policy
and DPA. For the authoritative, firm-level sub-processor list or a copy of our DPA, contact
privacy@4degrees.ai.
Infrastructure #
| Sub-processor | Purpose | Data processed | Location |
|---|---|---|---|
| Google Cloud Platform (Google LLC) | Application hosting (Cloud Run), primary database (Cloud SQL for PostgreSQL), managed cache (Memorystore), networking. All 4Degrees data is stored here, encrypted at rest. | CRM records, Slack workspace/user identifiers, encrypted Slack bot tokens, request metadata | United States (us-central1); Saudi Arabia (me-central2) |
AI / language-model processing #
When you ask the assistant a question, the relevant CRM data your seat is permitted to see is sent to a large language model to generate the answer. We route model calls through OpenRouter, which forwards the request to the upstream model provider.
| Sub-processor | Purpose | Data processed |
|---|---|---|
| OpenRouter, Inc. | LLM API gateway — routes each assistant request to the upstream model provider below. | The prompt for a single turn: the user’s question and the CRM data needed to answer it |
| Anthropic, PBC (Claude) | Primary language model that generates assistant answers. | Per-request prompt only; not retained for training (see below) |
| Google LLC (Gemini) | Fallback language model used if the primary model is unavailable. | Per-request prompt only; not retained for training (see below) |
Per their respective API and connector terms, the model providers do not train their models on data submitted through the API. Prompts are processed transiently to produce the answer and are not used to improve any model.
Error monitoring #
| Sub-processor | Purpose | Data processed |
|---|---|---|
| Functional Software, Inc. (Sentry) | Application error and performance monitoring. | Error events with numeric user/organization identifiers and request metadata. We do not send message content, tool arguments, tool results, or bot reply text to Sentry. |
What we do not share #
- We do not sell or rent your data to anyone.
- We do not share Slack or CRM data with third parties other than the sub-processors listed above and the AI assistant you yourself connect (Claude, ChatGPT) during a connector call.
- We do not use Slack or CRM data to train any AI model.
- Email and meeting body content is encrypted with per-tenant keys and is not readable by, or sent to, the connector, the Slack bot, or any language-model sub-processor — only metadata is.
Changes & notification #
We update this list when we add or replace a sub-processor for the connector or Slack app. Customers with a Data Processing Agreement can request advance notification of new sub-processors as provided in that agreement; contact privacy@4degrees.ai to opt in.
Contact #
Questions about sub-processors, our DPA, or data processing: privacy@4degrees.ai.