AI features
Echo's AI assists help you understand a response or prepare work you would otherwise write by hand. Open an assist when you need it, inspect the result and decide what to keep. Requests, collections, protocols and ordinary testing work without AI.
Choose an assist
| What you need | What Echo can help prepare |
|---|---|
| Understand a response | An explanation of the response and its contents |
| Investigate a failed call | Possible causes and steps to check |
| Start from a specification | A draft request from pasted API information |
| Import a cURL command | A reconstructed request to review |
| Configure authentication | A suggested authentication pattern |
| Reuse values from JSON | Suggested variable extractions |
| Test a working request | Assertions for its post-request script |
| Cover a collection | A proposed test suite |
| Prepare a dependency | Mock routes from a schema |
| Explain a request sequence | Flow documentation |
| Improve a chain | Proposed changes to its steps |
Authentication suggestions do not obtain credentials for you. Check the API's requirements and supply the correct values yourself. Review extracted variable names and destination environments before applying them. Generated assertions, requests and mocks need the same review as anything you write by hand.
Start from OpenAPI
The specification assistant combines a normal import with optional AI work:
- Choose the destination space and open or paste the OpenAPI document.
- Select the optional test, desktop-mock and flow-documentation steps you need.
- Review the collection preview and estimated token use.
- Start the selected steps and inspect their individual results.
The import does not use AI. Optional generation needs AI access and quota, and one step can be refused even if another succeeds. Work saved before a later failure or cancellation remains available.
Flow documentation is saved to DocuVault using your existing access. That does not create a public documentation link. A generated desktop mock also remains separate from public mock hosting.
Access and availability
AI requires a Pro AI seat assigned to your account, the organization's AI features enabled and sufficient quota. An organization or platform restriction can make a particular assist unavailable. If access is denied, check the reason shown and ask your organization administrator about the seat or policy.
Having access to one assist does not guarantee that every generation step will run. The interface's availability and quota messages describe the current request. You can continue using the non-AI editor, import and runner.
Review what is sent
Before sending AI context, Echo applies masking rules to known secret values, authentication fields, sensitive headers and supported body content. This reduces exposure, but it does not recognize every confidential fact that someone might paste into a specification, description or payload.
Review the input, remove information the assist does not need and use invented data in examples. AI needs a network connection and processes the supplied context through Gapfy's AI service. The encrypted database and guest mode do not make an AI request an offline operation.
Apply the result deliberately
Read a suggested request's URL, method, body and authentication before sending it. Check that an assertion expresses your API's real contract, rather than merely copying one observed response. For mocks, verify error cases as well as successful examples. For chain changes, review ordering, conditions and extracted values before applying the proposed plan.
Use automation and testing to run the reviewed checks, and environments and variables to keep reusable values outside request text.