feat(gateway): add interactive invoke console by aidandaly24 · Pull Request #1978 · aws/agentcore-cli

aidandaly24 · 2026-08-11T23:16:44Z

Summary

add a persistent POST JSON console for agentcore gateway invoke
add Gateway switching, raw path editing, generated/continued sessions, and live response streaming
preserve the headless Gateway invoke contract while allowing CLI flags to seed interactive context
share the multiline editor with Runtime and keep long payloads within stable terminal rows

UX

Ctrl+T switches Gateway and clears Gateway-specific state
Ctrl+P edits the Gateway-relative path; blank uses the exact Gateway URL
Ctrl+V toggles completed JSON between raw and pretty forms
CUSTOM_JWT Gateways proactively require a CLI-provided bearer token
Runtime and MCP session metadata continues across requests

Verification

206 pass, 0 fail focused Gateway/Runtime invoke tests
1197 pass, 0 fail full bun test src
bun run typecheck
bun run lint:check
bun run format:check
bun run build
live TUI validation: NONE, AWS_IAM, CUSTOM_JWT, AUTHENTICATE_ONLY, JWT passthrough, OAuth 3LO, OAuth OBO, inference, Runtime HTTP, MCP aggregation, Lambda, and real SSE streaming

Stack

Depends on #1967. Review this PR against feat/gateway-invoke.

codecov-commenter · 2026-08-13T19:46:16Z

Codecov Report

❌ Patch coverage is 98.43505% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 97.00%. Comparing base (ea3fa84) to head (6489a5d).

Files with missing lines	Patch %	Lines
src/handlers/gateway/invoke/screen.tsx	97.80%	10 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##           refactor    #1978      +/-   ##
============================================
+ Coverage     96.97%   97.00%   +0.02%     
============================================
  Files           364      366       +2     
  Lines         20844    21373     +529     
============================================
+ Hits          20213    20732     +519     
- Misses          631      641      +10

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

nborges-aws · 2026-08-13T20:29:49Z

+        metadata: metadata(response),
+        state: "streaming",
+      });
+      const kind = classifyStreamingResponse(response.contentType);


I think we need the same statusCode !== 204 && statusCode !== 205 clause here that we did in the headless invoke PR

Good catch. Fixed in e14e8848 by exempting HTTP 204 and 205 from the TUI binary/unknown-content guard, matching the headless invoke behavior. I also added screen coverage for both statuses with an empty content type and body; the focused Gateway invoke screen suite passes 31/31.

nborges-aws

LGTM

Hweinstock · 2026-08-14T16:42:36Z

      }

      const jsonOutput = ctx.require(JsonKey);
+      if (flags.payload === undefined) {


does this break the pattern of tui only no flags? If so, could this be confusing for customers?

Yes, but Runtime invoke already follows this pattern where a missing --payload launches the TUI. I think it technically could be a little bit confusing, but I personally think there's are a couple good reasons for why we should keep it this way. For one, we wanted to keep the TUI simple to start. When building runtime invoke I built an advanced options menu, which was accessible in the invoke TUI via a ctrl + key command. This added complexity to the code and we thought it would be better to start simple and layer this complexity on top in the future. So to keep these configurations accessible, I opted to not open the TUI on no --payload. Second, I think for things like --session-id it is more intuitive to pass it in via the CLI before opening the agent than after opening up the agent.

The harness invoke TUI is also not strictly no-flags only so I think it's okay to break the pattern on invoke for now.

I think it'd be worthwhile though to have a conversation about this with the team though.

Hweinstock · 2026-08-14T16:43:36Z

+              io,
+            );
+          } catch (error) {
+            if (error instanceof InvalidEnvironmentError) {


why do we need to re-wrap this? Is there a specific InvalidEnvironmentError that should be a InputValidationError?

Tbh this only exists to change the exit code. I agree with what I think you may be getting at which is that this is unnecessary complexity. I'll remove this.

Runtime invoke also uses this, and I'll update it there as well. I think what I could do is have this here throw a new InvalidEnvironmentError(message, { exitCode: ExitCode.USAGE }) instead.

https://github.com/aws/agentcore-cli/blob/refactor/src/tui/index.tsx#L49-L51

Then remove the catch/rewrite here and in Runtime.

Hweinstock · 2026-08-14T16:46:57Z

+  test("deep-links an id-only invoke and seeds interactive request context", async () => {
+    const core = configuredCore();
+    const output = captureIO();
+    const render = spyOn(tui, "renderTuiAt").mockResolvedValue(undefined);


is there a way to avoid spying by leveraging the existing dependency injection?

Hweinstock · 2026-08-14T16:48:19Z

+
+test("preserves unexpected TUI bearer-token source failures", async () => {
+  const failure = new TypeError("source failed");
+  const resolve = spyOn(SourceResolver.prototype, "resolveText").mockRejectedValue(failure);


same question as above. Is there way we can inject real implementations?

Hweinstock · 2026-08-14T16:49:51Z

+): Promise<string | undefined> {
+  if (source === "-") {
+    throw new InputValidationError(
+      "stdin bearer tokens are not available when launching the interactive console",


don't we need to support stdin for sensitive fields?

This is unfortunately a trade off of using the TUI. Headless invoke supports --bearer-token -. Interactive invoke deliberately does not because Ink needs the same stdin for keyboard input. Consuming it for the token would drain or block the TUI. Interactive mode supports inline and file:// tokens. So you still can pass bearer token securely through a file for interactive. Runtime has the same restriction.

Hweinstock · 2026-08-14T16:50:12Z

+  try {
+    return await new SourceResolver({ stdin }).resolveText("bearer-token", source);
+  } catch (error) {
+    if (error instanceof SourceResolutionError) {


whats wrong with the sourceResolutionError bubbling up?

Hweinstock

lgtm! thanks for answering my earlier questions, those decisions make sense.

github-actions Bot added the size/xl PR size: XL label Aug 11, 2026

aidandaly24 had a problem deploying to e2e-testing August 11, 2026 23:17 — with GitHub Actions Failure

github-actions Bot added agentcore-harness-reviewing AgentCore Harness review in progress size/xl PR size: XL and removed agentcore-harness-reviewing AgentCore Harness review in progress size/xl PR size: XL labels Aug 11, 2026

aidandaly24 had a problem deploying to e2e-testing August 12, 2026 01:50 — with GitHub Actions Failure

github-actions Bot removed the size/xl PR size: XL label Aug 12, 2026

aidandaly24 had a problem deploying to e2e-testing August 12, 2026 17:51 — with GitHub Actions Failure

github-actions Bot added the size/xl PR size: XL label Aug 12, 2026

aidandaly24 force-pushed the feat/gateway-invoke-tui branch from 196b394 to 9c11253 Compare August 13, 2026 18:44

github-actions Bot added size/xl PR size: XL and removed size/xl PR size: XL labels Aug 13, 2026

aidandaly24 had a problem deploying to e2e-testing August 13, 2026 18:45 — with GitHub Actions Failure

github-actions Bot added size/xl PR size: XL and removed size/xl PR size: XL labels Aug 13, 2026

aidandaly24 had a problem deploying to e2e-testing August 13, 2026 19:16 — with GitHub Actions Failure

aidandaly24 force-pushed the feat/gateway-invoke-tui branch from d312820 to ad207ae Compare August 13, 2026 19:41

Base automatically changed from feat/gateway-invoke to refactor August 13, 2026 19:43

aidandaly24 force-pushed the feat/gateway-invoke-tui branch from ad207ae to 858db4d Compare August 13, 2026 19:43

aidandaly24 marked this pull request as ready for review August 13, 2026 19:45

nborges-aws reviewed Aug 13, 2026

View reviewed changes

aidandaly24 force-pushed the feat/gateway-invoke-tui branch from 858db4d to 40b3457 Compare August 13, 2026 20:31

aidandaly24 requested a review from nborges-aws August 13, 2026 20:35

nborges-aws previously approved these changes Aug 13, 2026

View reviewed changes

aidandaly24 requested review from jariy17 and tejaskash August 13, 2026 20:37

Hweinstock reviewed Aug 14, 2026

View reviewed changes

aidandaly24 added 6 commits August 14, 2026 17:23

feat(gateway): add interactive invoke console

b0b5893

fix(tui): keep long multiline input within its rows

900d0af

fix(gateway): refine path editor and session labels

a536851

style(gateway): use neutral path dialog border

cad95f3

fix(gateway): make idle escape navigate back

418fe96

fix(gateway): allow no-content responses in TUI

fdbfe3a

aidandaly24 force-pushed the feat/gateway-invoke-tui branch from e14e884 to fdbfe3a Compare August 14, 2026 17:23

refactor(invoke): simplify TUI launch errors

3e0bb88

aidandaly24 dismissed nborges-aws’s stale review via 3e0bb88 August 14, 2026 17:27

aidandaly24 requested a review from Hweinstock August 14, 2026 17:27

aidandaly24 added 2 commits August 14, 2026 17:29

test(gateway): allow TUI startup on slow runners

53ba0d3

test(gateway): inject invoke TUI renderer

6489a5d

Hweinstock approved these changes Aug 14, 2026

View reviewed changes

nborges-aws approved these changes Aug 14, 2026

View reviewed changes

aidandaly24 merged commit 1976766 into refactor Aug 14, 2026
11 checks passed

aidandaly24 deleted the feat/gateway-invoke-tui branch August 14, 2026 21:05

Uh oh!

Conversation

aidandaly24 commented Aug 11, 2026

Summary

UX

Verification

Stack

Uh oh!

codecov-commenter commented Aug 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

nborges-aws left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Hweinstock left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

codecov-commenter commented Aug 13, 2026 •

edited

Loading