> For the complete documentation index, see [llms.txt](https://docs.pharos.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.pharos.ai/configuration-guide/security.md).

# Security & Data Access

What data stays in your org, what each integration sends outside it, and which permissions Pharos requires — the answers a security review asks for.

Pharos runs entirely inside your Salesforce org. Logs, issues, rules, and settings are native custom objects, so your observability data inherits your org's security model rather than being copied to a vendor platform. See [Native Data Model](https://docs.pharos.ai/platform-overview/pharos-platform-architecture/logs/log-storage/native-data-model.md) for the storage architecture.

Data only leaves your org when **you** connect an integration and configure a rule that sends something. This page summarizes exactly what that means per integration — the question most information security reviews start with.

***

## What Each Integration Sends Outside Your Org

| Integration                                                                                 | Authentication                                                                                                                     | What leaves your org                                                                                                                                                                                                                                                                     | Access required in the external system                                                                                                                                                                            |
| ------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| [Slack](https://docs.pharos.ai/configuration-guide/integrations/slack.md)                   | OAuth — the Pharos app or your own [custom app](https://docs.pharos.ai/configuration-guide/integrations/slack/slack-custom-app.md) | Notification content only: whatever your message templates include (typically issue summary, category, impact stats, and a link back to Pharos)                                                                                                                                          | Seven scopes: `chat:write`, `chat:write.public`, `channels:read`, `groups:read`, `users:read`, `im:read`, `im:write`                                                                                              |
| [Teams](https://docs.pharos.ai/configuration-guide/integrations/teams.md)                   | OAuth via Microsoft Entra ID against a Pharos-published verified app — no app registration needed                                  | Notification content only, as above                                                                                                                                                                                                                                                      | Entra ID consent. Depending on tenant policy this is granted by an admin directly, by a user consenting on the organization's behalf, or through an admin consent request                                         |
| [PagerDuty](https://docs.pharos.ai/configuration-guide/integrations/pagerduty.md)           | API token plus a Service Integration Key (Events API v2)                                                                           | Incident content: summary, severity, and a link back to Pharos                                                                                                                                                                                                                           | An API user token or general-access REST key, and an escalation policy on the target service                                                                                                                      |
| [Jira](https://docs.pharos.ai/configuration-guide/integrations/jira.md)                     | OAuth for Jira Cloud; API token for Server / Data Center                                                                           | Ticket content: whatever your ticketing templates include, plus comments on subsequent updates                                                                                                                                                                                           | Permission to create and edit issues in the target project                                                                                                                                                        |
| [Azure DevOps](https://docs.pharos.ai/configuration-guide/integrations/azure.md)            | OAuth                                                                                                                              | Work item content, as above                                                                                                                                                                                                                                                              | Permission to create and edit work items; third-party OAuth access must be enabled on the organization                                                                                                            |
| [GitHub](https://docs.pharos.ai/configuration-guide/integrations/github.md)                 | OAuth or API key                                                                                                                   | Org **metadata**, pushed as auto-merged pull requests: Apex classes and triggers, flows, workflow rules, objects and fields, Aura and Visualforce components, custom metadata types, remote site settings, approval processes, **connected apps, and named credentials**. No record data | Access to the target repository. Use a private, dedicated repo                                                                                                                                                    |
| [OpenAI](https://docs.pharos.ai/configuration-guide/integrations/openai.md)                 | API key                                                                                                                            | Error messages, stack traces, log metadata, metadata structure, aggregated statistics, and audit trail metadata. See below                                                                                                                                                               | An OpenAI account with billing configured                                                                                                                                                                         |
| [AI Agents & MCP](https://docs.pharos.ai/configuration-guide/integrations/ai-agents-mcp.md) | The `@pharosai/mcp` server plus a Pharos permission set on the connected user                                                      | Whatever the agent reads in response to your queries, bounded by the permission set you assign                                                                                                                                                                                           | None beyond the agent itself. The server runs locally and opens no network port — see [How the Server Connects](https://docs.pharos.ai/configuration-guide/integrations/ai-agents-mcp.md#how-the-server-connects) |

{% hint style="info" %}
None of the integrations above is enabled by default. Until you connect one and configure a rule or action that uses it, nothing is sent to it.

Email is the exception worth naming: it needs no integration because it uses Salesforce's own delivery. An email notification action therefore sends content out of the org — through your existing Salesforce email infrastructure, under your existing controls — without anything being connected first.
{% endhint %}

***

## AI Data Handling

OpenAI is the integration that draws the most scrutiny, so it is worth stating precisely.

**Sent to OpenAI:** error messages, stack traces, and log metadata; metadata structure (flows, triggers, Apex classes, validation rules); aggregated statistics (occurrence counts, timestamps, user impact); and audit trail metadata (who changed what, when).

**Not sent:** raw record data from Salesforce objects, customer or business transaction details, user-identifiable information beyond error contexts, and sensitive or encrypted fields.

Two things enforce that boundary rather than merely describing it. Pharos's trust layer governs what may cross at all, and [Sensitive Data Masking](https://docs.pharos.ai/configuration-guide/advanced/sensitive-data-masking.md) removes matched values *before* logs are persisted — so anything your masking rules catch is already redacted in storage and cannot reach OpenAI even in an error message or stack trace.

Three further guarantees matter for a security review:

* **Your data is never used to train external AI models.**
* **The vector index is stored in vectorized form** — mathematical representations that are extremely difficult to reverse-engineer even if accessed.
* **AI analysis respects your existing Salesforce security model.** Sharing rules, profiles, and permission sets all apply, so users only see analysis for data they can already access. There is no separate access-control system.

For why this boundary is drawn where it is, see [The Pharos Trust Layer](https://docs.pharos.ai/platform-overview/pharos-platform-architecture/ai-analysis/issue-analysis.md#the-pharos-trust-layer).

***

## Controlling What Gets Captured

Two controls limit sensitive data before it is ever stored or forwarded:

* [**Sensitive Data Masking**](https://docs.pharos.ai/configuration-guide/advanced/sensitive-data-masking.md) — regex rules redact matched values *before* logs are persisted. Redaction is irreversible, and masked values stay masked everywhere the log appears, including exports and notifications.
* [**Log Retention**](https://docs.pharos.ai/configuration-guide/log-retention.md) — age-based rules remove or archive log records so sensitive diagnostic detail is not retained longer than your policy allows.

For multi-org setups, [Connected Orgs](https://docs.pharos.ai/configuration-guide/connected-orgs.md) controls direction explicitly: each connection is configured to send to, or receive from, a given org, with optional filter criteria narrowing what is shared.

***

## Access Control Inside Salesforce

Pharos ships three permission sets — `Log_Read`, `Log_Read_Write`, and `Log_Admin` — distributed separately from the managed package as the [pharos-permissions](https://resources.pharos.ai/pharos-permissions) download, because they include user permissions Salesforce does not allow inside a managed package.

Because Pharos data is native, standard Salesforce controls apply without extra configuration: field-level security, sharing rules, object CRUD permissions, and platform encryption. See [Access Controls](https://docs.pharos.ai/platform-overview/pharos-platform-architecture/logs/access-controls.md) for the model and [Pharos Running User](https://docs.pharos.ai/configuration-guide/the-basics/pharos-running-user.md) for the identity background operations execute under.

***

## Related

* [Pharos Running User](https://docs.pharos.ai/configuration-guide/the-basics/pharos-running-user.md) — the identity Pharos uses for background processing
* [Sensitive Data Masking](https://docs.pharos.ai/configuration-guide/advanced/sensitive-data-masking.md) — redact values before storage
* [Integrations](https://docs.pharos.ai/configuration-guide/integrations.md) — connection setup for each system
* [Access Controls](https://docs.pharos.ai/platform-overview/pharos-platform-architecture/logs/access-controls.md) — permission set model and data protection
