OpenAI AI agent integration

What you can do with the OpenAI integration

Give your AI agents authenticated access to OpenAI so they can read, act, and update records on your behalf — inside a governed, observable workflow.

• Let agents read data from OpenAI as part of a workflow.
• Let agents take typed actions inside OpenAI with allow/deny policies.
• Combine OpenAI with other tools to automate cross-system workflows.
• Audit every tool call in Monitor — agent, action, input, response, latency, cost.

How to connect OpenAI

Paste your API key into the credential form. Dezifi encrypts it at rest and never logs it.

1. 1

   Open Integrations in your workspace

   Navigate to Integrations from the left sidebar. The catalog opens to all available tools.
2. 2

   Find OpenAI

   Search the catalog or filter to the AI & Machine Learning category. Click the OpenAI card.
3. 3

   Connect your account

   Click Connect. Paste your API key into the credential form. Dezifi encrypts it at rest and never logs it.
4. 4

   Test the connection

   Run the built-in connection test. Dezifi calls a low-impact endpoint and confirms the credential works.
5. 5

   Grant access to an agent

   Open an agent in the builder, go to Tool Selection, check OpenAI, save. The agent can now invoke OpenAI actions during runs.

Governance and safety

OpenAI is a tool like any other in Dezifi — its actions are governed by your policies, inspected by guardrails, and recorded in every run trace.

• Tool allow / deny — restrict which agents can invoke which actions.
• Approval gates — require human sign-off before destructive operations.
• Rate limits — cap how often an agent can call the integration.
• Audit trail — every call, input, response, and identity is logged.

Frequently asked questions

How do I connect OpenAI to Dezifi?

Open Integrations in your workspace, search for OpenAI, click Connect, and follow the api key flow. Once connected, the integration becomes available as a tool for any agent.

Which AI agents can use the OpenAI integration?

Any agent in your Dezifi workspace can be granted access to OpenAI. You scope which agents see it via tool allow-lists on policies.

Is the OpenAI integration secure?

Yes. Credentials are encrypted at rest, scoped to your workspace, and never shared across tenants. Every tool call is logged in the run trace.

Can I limit what the agent can do in OpenAI?

Yes. Policies let you allow or deny specific actions and require human approval for high-risk operations. The principle of least privilege is enforced at runtime.

Other AI & Machine Learning integrations