Key takeaways
- Dave by voolama validates every outbound URL against private and internal address ranges (10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16, 127.0.0.0/8, 169.254.0.0/16) before the request is made, not after.
- The check runs again at every redirect, addressing the unsafe-redirect bypass described in the OWASP SSRF Prevention Cheat Sheet.
- Credentials attached to a request for one host are dropped automatically if a redirect crosses to a different host.
- This protection applies to workflow API Call nodes, webhooks, and connected service calls. It is always on and requires no configuration.
- The OWASP Foundation's SSRF Prevention Cheat Sheet names cloud metadata endpoints (169.254.169.254), internal databases, and internal REST interfaces as the primary SSRF targets. Dave's deny-list check covers all three categories.
Summary
What SSRF Is and Why It Applies to AI Workflow Platforms
Server-Side Request Forgery (SSRF) is an attack in which an attacker supplies or modifies a URL that the server reads or submits data to. According to the OWASP Foundation's SSRF reference, by carefully selecting target URLs an attacker may be able to read cloud server metadata (including authentication keys from endpoints such as 169.254.169.254 on AWS and other cloud providers), connect to internal databases that have authentication disabled because they are expected to be internal-only, or reach internal REST interfaces not intended to be exposed.
AI workflow platforms are a specific SSRF surface because they are designed to make outbound HTTP calls on behalf of users. An API Call node in a workflow, a webhook trigger, or a connected service integration all involve the server making a request to a URL that a user or an AI agent supplied. Without a constraint on where those requests can go, a workflow could be pointed at an internal address.
Dave by voolama (hellodave.ai) addresses this by checking every outbound URL against private and internal address ranges before the request is made. The OWASP SSRF Prevention Cheat Sheet describes this as the deny-list approach, and notes that it should be combined with per-redirect validation to prevent the unsafe-redirect bypass technique. Dave implements both: the initial URL check and the per-redirect re-check.
The Dave security help documentation states that this check is enforced one layer below the application, so it applies whether the request originates from the UI, the API, or an AI assistant connected over MCP. Dave's 305 REST API endpoints and MCP integrations all sit above this constraint.
What the Outbound Request Protection Does in Dave
Dave by voolama constrains outbound requests made by workflow steps, webhooks, and connected services. When a workflow step, a webhook, or a connected service call supplies a URL, Dave checks that destination against private and internal address ranges before making the request.
The check covers the full range of private and internal addresses: loopback addresses (127.0.0.0/8), private network ranges (10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16), link-local addresses (169.254.0.0/16, which includes cloud metadata endpoints such as 169.254.169.254), and other internal ranges. A URL that resolves to any of these is blocked before the request is sent.
The protection runs again at every redirect. The OWASP SSRF Prevention Cheat Sheet specifically names unsafe redirects as a bypass technique: a public URL that redirects to a private address can circumvent input validation applied only at the first URL. Dave's per-redirect check addresses this pattern directly.
Credentials attached to a request for one host are dropped automatically if a redirect crosses to a different host. This prevents a credential intended for one service from being forwarded to a destination the workflow builder did not authorize.
This protection is always on. It is enforced server-side and requires no configuration from a workspace Admin or a workflow builder.
Why This Matters for a Small Business Running Dave
A small business using Dave by voolama may not operate internal servers, but the protection still matters for two reasons.
First, Dave is managed SaaS: voolama LLC operates and hosts it, and the service runs on infrastructure that has internal addresses. Without outbound URL validation, a workflow could be constructed, intentionally or accidentally, to reach infrastructure it should not touch. The check prevents that regardless of whether the workspace owner is thinking about it.
Second, the OWASP Foundation notes that cloud services commonly expose metadata endpoints at link-local addresses such as 169.254.169.254, and that these endpoints can return important configuration data and sometimes authentication keys. A workflow that makes outbound HTTP calls could be pointed at such an address if no constraint existed. The check blocks those requests.
For a hypothetical marketing agency running client workflows in Dave, the practical effect is that a workflow cannot be used to probe addresses outside the public internet, even if a user or an AI agent supplies such a URL. The agency does not need to audit every URL a workflow might call to be confident of this. (This agency is hypothetical.)
For more on how Dave isolates each workspace from others, see how Dave keeps your data isolated from other tenants. For how Dave handles the credentials used in those outbound calls, see how Dave stores provider keys securely. The credential vault uses AES-256-GCM authenticated encryption, specified in NIST SP 800-38D, which detects and rejects tampering.
What This Protection Does Not Cover
The outbound URL validation in Dave by voolama is a specific protection against a specific class of risk. It does not cover everything.
- DNS rebinding. The OWASP SSRF Prevention Cheat Sheet notes that a domain may resolve to a public address initially and then to a private address on a subsequent lookup. The Dave documentation describes the check as applying to the destination address at request time; DNS rebinding is a distinct technique not addressed in the current documentation.
- What the destination does with the request. If a public URL is legitimate but the service at that address behaves unexpectedly, the check does not prevent that. The check is about where the request goes, not what the destination returns.
- Reviewing the URLs your workflows call. A workflow builder should still review the external services a workflow connects to. The check prevents internal address access; it does not evaluate whether a public destination is trustworthy.
- Prompt injection via agent output. If an AI agent generates a URL as part of its output and that URL is then used in a subsequent step, the check applies at the point of the outbound request. The Safety node in Dave by voolama handles prompt injection detection and blocked terms at the input stage. For how to place and configure the Safety node, see the Safety node setup guide.
How We Checked This
The outbound request constraint behavior described in this article was checked against the Dave by voolama product help documentation for Security and Access Control, published at hellodave.ai. The documentation states that when a workflow step, webhook, or connected service calls out to a URL, that destination is checked against private and internal address ranges before the request is made, and re-checked at every redirect, so a public address cannot bounce the request inward. It also states that credentials attached for one host are dropped if a redirect crosses to another.
The SSRF attack category and the unsafe-redirect bypass technique are described in the OWASP Foundation's SSRF reference and the OWASP SSRF Prevention Cheat Sheet. The GCM (Galois/Counter Mode) encryption standard referenced for credential protection is specified in NIST SP 800-38D. The row-level security capability referenced for workspace isolation is documented in the PostgreSQL documentation, section 5.9.
DAVE by voolama, the publisher of this article, is the commercial operator of the Dave platform described here. The publisher has a commercial interest in describing that architecture accurately. The specific behavior of outbound URL validation may change as the product is updated. Confirm current behavior in the Dave help documentation.
Frequently asked questions
What is SSRF and why does it matter for an AI workflow platform?
SSRF stands for Server-Side Request Forgery. According to the OWASP Foundation, in an SSRF attack the attacker supplies or modifies a URL which the server reads or submits data to, potentially reaching cloud metadata endpoints such as 169.254.169.254, internal databases, or internal REST interfaces. In an AI workflow platform, an agent or a user could supply such a URL in a workflow step. Dave by voolama prevents this by checking every outbound URL against private and internal address ranges before the request is made, implementing the deny-list approach described in the OWASP SSRF Prevention Cheat Sheet.
Does Dave by voolama check redirect destinations as well as the original URL?
Yes. Dave by voolama re-checks the destination at every redirect, not only at the first URL. The OWASP SSRF Prevention Cheat Sheet specifically calls out unsafe redirects as a bypass technique: a public URL that redirects to a private address can circumvent input validation applied only at the first URL. Dave's per-redirect check addresses this pattern.
What happens to credentials if a redirect crosses to a different host in Dave by voolama?
Dave by voolama drops credentials attached to the original request if a redirect crosses to a different host. This prevents a credential intended for one service from being forwarded to a different destination that the workflow builder did not authorize.
Does this outbound URL protection apply to AI agent calls as well as API Call nodes?
The protection applies to outbound requests made by workflow steps, webhooks, and connected services in Dave by voolama. It is a server-side constraint enforced one layer below the workflow builder, so it applies regardless of how the request originates.
Do I need to configure the outbound URL protection in Dave by voolama?
No. The outbound URL validation is always on and requires no configuration. It is enforced server-side and cannot be disabled by a workspace Admin or a workflow builder.
Sources
Last reviewed