Debugging multi-step gRPC flows no longer means copy-pasting tokens, user IDs, and trace IDs between calls. RPCORA is a local-first gRPC chain debugging workbench: define a call, link calls into a chain, and each response feeds the next request.
{
"username": "admin",
"password": "admin123"
}{
"token": "jwt.user-1.1777898990",
"expires_in": "86400",
"user_id": "user-1"
}The real RPCORA UI: the token returned by Login becomes the input of the next request.
Postman, Insomnia, or grpcurl can all fire a single request — but real business debugging is never a single request.
Tokens, user IDs, and trace IDs from a login response are pasted into the next request by hand — slow and error-prone.
An extractor stores body.token as session.token; downstream requests reference it as {{session.token}}.
Every regression means manually replaying login, query, and create — and remembering what changed.
Save working requests as a chain, run them in order, and see exactly which step failed.
In generic tools, scenarios live in bookmarks and chat logs — a new owner has to start from zero.
Requests, environments, and chains stay in a local workspace; failures produce a redacted Doctor report.
A real business chain as the example: log in for a token → fetch the user with it → create an order with the user ID.
Pick a method from the API tree built by Reflection or Proto, then fill the payload from schema.
Declare an extractor on the upstream step: body.token → session.token.
Downstream requests reference upstream results as {{session.token}}, connecting login, query, and create.
Re-run the whole chain any time; failed steps produce Doctor evidence automatically.
username · passwordbody.token → session.tokenauthorization: {{session.token}}body.user_id → profile.userIduser_id: {{profile.userId}}body.order_id → order.idGeneric tools can fire a single gRPC request. RPCORA is different because multi-step calls are first-class citizens.
Extractors, runtime variables, and multi-step replay are the main path, not a plugin or a script. Login, auth, query, and create become one repeatable chain.
Chains built-inWorkspaces, schemas, responses, and chain context are stored locally. No account, no cloud sync — safe for internal networks and sensitive environments.
No account · No cloudChain steps, gRPC status, duration, schema source, and Doctor results form a redacted report you can hand to another engineer.
Doctor report| RPCORA | Generic gRPC clients | Scripts · CLI | |
|---|---|---|---|
| Context between calls | ✓Responses feed the next call | Variables + scripts by hand | Copy-paste |
| Multi-step chain replay | ✓Re-run the whole chain in one click | Depends on runner / test scenarios | Redo every step |
| Local privacy + debug evidence | ✓Local storage + redacted reports | Cloud sync first | None |
Connect to a service, read schema, compose Payload, Metadata, and Auth, then run the request and inspect the response. When Reflection is unavailable, import a Proto folder and keep debugging.
When a call fails, RPCORA keeps resolved context, Doctor checks, and chain reports in one evidence panel. You can see where the failure happened, which context was present, and what to inspect next.
{{session.token}}, {{profile.user_id}}, trace IDs, and other context are visible before runtime.
Context readyTarget, TLS, Metadata, Reflection, Proto fallback, and gRPC status stay separated.
4 checks passedReports keep chain steps, duration, error summary, and next actions while hiding sensitive values.
RedactedRun a single gRPC method, save reusable requests, and replay the context behind authenticated business calls.
Fix login, query, create, and verify steps into a chain, then share a redacted failure report with engineers.
Quickly separate target, certificate, permission, reflection, Proto, and response-shape failures.
Use it for desktop app feedback, gRPC workflow questions, and Proto, Reflection, TLS, or Auth debugging cases.
Scan with QQ to join
Start small: install the desktop app, run one request in APIs, save it, then replay a two-step chain.
For Apple Silicon and Intel Mac. Recommended for regular installation.
For Windows x64 desktops with the standard installer flow.
Single-file AppImage package. Make it executable before launching.