SOC 2 for SaaS Integrations: Scope OAuth Grants and Evidence
Map customer-connected SaaS integrations for SOC 2. Track OAuth grants, data flows, webhooks, evidence sources, and the controls your team operates.

For SOC 2, a SaaS integration needs a clear place in your service boundary. List each connection that reads, writes, or can reach customer data, then record who authorizes it, what access it gets, where data moves, and which system can prove your controls ran. An OAuth consent screen or a provider’s SOC 2 report cannot answer all of those questions for your product. Start with an integration inventory before you collect audit screenshots.
When is a SaaS integration in SOC 2 scope?
The AICPA Trust Services Criteria provide criteria for evaluating controls over the system management describes. They do not contain a universal list of SaaS integrations or prescribe an OAuth implementation. Management describes the service, commitments, risks, and controls. The independent CPA firm performs the examination and decides what evidence is sufficient for the agreed scope.
Include the integration in your scope discussion when it helps deliver the service in the report, handles in-scope data, or can affect an in-scope control. For example, a CRM sync that receives customer records, a repository app that can read source code, and a webhook receiver that changes customer state each create different paths to map. An app installed only for internal convenience may have a different role, but its access and vendor risk still deserve review. Use the SOC 2 scope guide for the whole service boundary and a data flow diagram for the actual information paths.
Build an inventory of connections, grants, and owners
Keep one record per provider or app integration and a way to identify its active customer connections in the source system. Do not put live credentials in the inventory. For each integration, capture:
| Field | Question to answer |
|---|---|
| Purpose and owner | Which feature uses the connection, and who maintains it? |
| Customer grant | Who can connect it, whose account grants access, and how is consent recorded? |
| Scope and data | Which API permissions, fields, files, and events can it reach? |
| Direction and storage | What enters or leaves your service, and where is a copy retained? |
| Runtime paths | Which token store, webhook endpoint, queue, worker, and support tool touch it? |
| End of access | What happens when a customer disconnects, an owner leaves, or a grant expires? |
| Source evidence | Which system has the configuration, authorization tests, grant list, logs, and change history? |
This is a decision record, not a claim that the requested scopes are safe. A team may find that a provider offers only a broad scope for a narrow feature. Record that limit, the data it exposes, the compensating controls you actually operate, and the owner who accepts or changes the design. If a new integration adds a data type or destination, update the scope and diagram before treating last quarter’s review as current.
Separate protocol design from SOC 2 evidence
RFC 9700, the OAuth 2.0 Security Best Current Practice, and the OWASP OAuth2 Cheat Sheet provide technical guidance for flows, redirect handling, token protection, and other OAuth risks. These are useful engineering sources, not a claim that SOC 2 requires every recommendation in every integration.
Map each control your team chose to a source record. For example:
| Control question | Possible source record |
|---|---|
| Does the app request only the scopes needed for its feature? | Provider app configuration and reviewed permission change |
| Is a customer connection bound to the right tenant? | Authorization test using two synthetic tenants and the deployed version |
| Can a webhook sender be verified before it changes state? | Receiver test, configuration, and rejected-event log |
| Can operators access a stored token? | Token-store role configuration and dated access review |
| What happens after disconnect or revocation? | Disconnect test, source grant state, and job result |
The source systems are where the controls run. Your GRC record should identify the source, owner, period, and review result without copying tokens or raw customer data into Git. Keep sensitive raw output in an access-controlled system and link to a fixed export or approved reference when needed. A dated control record can then show which evidence was reviewed and what gaps remain.
Test a customer disconnect from end to end
A connection may disappear from the UI while a refresh token, queued job, or webhook route still works. Test the whole path with synthetic customer data before promising that disconnect removes access:
- Create a test connection and record the granted scopes and the source connection ID, without saving token values in the test record.
- Run one authorized sync and record the expected data movement and job ID.
- Disconnect the integration as the customer. Check the application state, provider grant state where visible, token handling, and queued jobs.
- Try the same sync or webhook path again. Record the denied or safely ignored result and check that no new customer data was written.
- Have an owner review the source output and any failed step. Fix and retest before claiming the behavior works.
This example test supports a narrow claim for the provider, flow, roles, and version exercised. Some providers do not expose grant state or offer the same revocation behavior. Record what you can verify directly, what the provider documents, and what remains unknown. A passing disconnect test does not prove that every existing customer connection has the same state. For a Type 2 examination, keep dated operating evidence for the controls in the specified period, including exceptions and follow-up.
Keep third-party and customer responsibilities explicit
The provider operates its own API, identity service, and logs. Your team chooses requested scopes, runs its code, protects credentials it holds, authorizes customer actions, and responds to failures under its commitments. Customers may control who installs the app and what data they authorize. Write those boundaries down; do not use a provider’s SOC 2 report as proof that your own integration controls operated. Use the vendor review guide for the provider’s role, report scope, exceptions, and follow-up. For customer-facing boundaries, the multi-tenant SaaS guide shows how to map shared authorization paths.
FileGRC is a Git-native GRC workspace for SOC 2 work. JSON holds structured records, Markdown holds long-form work, and Git supplies the change history. In the FileGRC workspace, a founder or engineer can keep the integration inventory, scope narrative, control records, evidence references, and review decisions together. An agent can inspect those current records and draft an auditor answer with sources and gaps flagged. Starter records remain proposals, and the application and provider systems still operate the controls. Management reviews the answer before sharing it; the independent CPA firm performs the examination and judges evidence sufficiency.
Run your SOC 2 program as files in Git.
Keep policies, controls, work, and evidence indexes in a repository your team and agents can inspect. Add optional FileGRC Autopilot email and Slack reminders to keep work moving.
Frequently asked questions
Are customer OAuth integrations in SOC 2 scope?
If an integration helps deliver the service in the report or can access in-scope customer data, describe its role, data flow, and relevant controls when defining the system boundary with the CPA firm. The exact scope depends on the service and commitments management describes, not on OAuth alone.
Does SOC 2 require a specific OAuth flow or token lifetime?
The AICPA Trust Services Criteria do not prescribe one OAuth flow or universal token lifetime for every SaaS integration. Choose controls that fit the service and its risks, use current OAuth security guidance when implementing them, and preserve evidence that the chosen controls operated.
What evidence should a SaaS integration keep for SOC 2?
Keep an inventory of connections and authorized scopes, source configuration or change records for the control, dated access and grant reviews, test results for authorization and disconnect behavior, and relevant incident or exception records. The CPA firm decides what evidence is sufficient for the scoped examination.
Is a provider's SOC 2 report enough for our integration?
No. A provider report can inform the team's vendor review, but it does not show that your application requested suitable scopes, protected its tokens, authorized customer actions, handled webhooks, or removed access when a connection ended. Document which controls each party operates.
Can FileGRC collect OAuth or webhook logs automatically?
Do not assume that. FileGRC is a Git-native GRC workspace for SOC 2 work. It can hold structured records in JSON, long-form work in Markdown, and Git change history, while application and provider systems operate controls and produce source logs. Your team reviews the evidence and an independent CPA firm performs the examination.