fix: detect claude CLI with inline args for MCP injection#584
Merged
khaliqgant merged 2 commits intomainfrom Mar 18, 2026
Merged
fix: detect claude CLI with inline args for MCP injection#584khaliqgant merged 2 commits intomainfrom
khaliqgant merged 2 commits intomainfrom
Conversation
![devin-ai-integration[bot]](https://avatars.githubusercontent.com/in/811515?s=60&v=4){kind=small}
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
When cli contains inline args (e.g. 'gemini --model foo'), Command::new(cli) fails because it treats the entire string as an executable path. Now extract just the binary via shlex::split before passing to Command::new and manual_cmd.
khaliqgant
added a commit
that referenced
this pull request
Mar 19, 2026
* fix: detect claude CLI with inline args for MCP injection * fix: extract executable from cli string in gemini/droid mcp setup When cli contains inline args (e.g. 'gemini --model foo'), Command::new(cli) fails because it treats the entire string as an executable path. Now extract just the binary via shlex::split before passing to Command::new and manual_cmd.
khaliqgant
added a commit
that referenced
this pull request
Mar 19, 2026
* feat: add workflow to polish CLI output with listr2 + chalk
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* ci: add workflow validation and dry-run check on PR
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(ci): fetch full history for git diff, run dry-run on all workflow types
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(sdk): polish workflow CLI output with listr2 spinners and chalk colors
- Replace plain console.log progress in cli.ts with listr2 task list
- Per-step spinners show owner, retry, nudge, force-release, and review events
- chalk colors: cyan for timestamps, green/red/yellow for status, dim for metadata
- logRunSummary() and broker stderr use chalk for visual hierarchy
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* test: add smoke test workflow for listr2 output rendering
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: use ESM-compatible import for check-sdk step
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(sdk): export createWorkflowRenderer for listr2 output in TS workflows
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: pre-attach catch to prevent unhandled rejection on fast-failing steps
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: use worker preset for verify step, mute console during listr rendering
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: filter [broker]/[workflow] noise, show observer URL, add unmount()
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: add missing unhandled-rejection guards and output filter to YAML path
- cli.ts: installOutputFilter() in runWithListr so YAML workflows also
suppress [broker]/[workflow HH:MM] noise during listr rendering
- cli.ts: done.catch()/workflowDone.catch() guards for fast-failing steps
- listr-renderer.ts: workflowDone.catch() guard for instant run:failed
- listr-renderer.ts: add renderer.unmount() to JSDoc example
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: address PR review feedback for workflow output polish
- Output filter: use regex .test() instead of startsWith/^ anchor so
chalk-colored [broker] and [workflow HH:MM] lines are properly suppressed
- Resume mode: add event listener for step progress reporting
- GHA workflow: fix grep exit code with || true, use env var instead of
raw ${{ }} interpolation (script injection), use npx tsx instead of
non-existent 'run' subcommand, only validate/dry-run YAML files
- Workflow: fix incorrect CJS assumption (SDK is ESM), add final
type-check gate after review step
* fix: CI failures and skipped-step visibility
- Add chalk and listr2 to root package.json (Build & Validate requires them)
- Dynamic import listr2 so SDK loads on Node 18 (styleText not available)
- Show steps skipped without prior start event in listr output
- Remove unused ListrType import
* fix: detect claude CLI with inline args for MCP injection (#584)
* fix: detect claude CLI with inline args for MCP injection
* fix: extract executable from cli string in gemini/droid mcp setup
When cli contains inline args (e.g. 'gemini --model foo'),
Command::new(cli) fails because it treats the entire string as
an executable path. Now extract just the binary via shlex::split
before passing to Command::new and manual_cmd.
* chore(release): v3.2.8
* bump versions (#590)
* bump versions
* fix: refresh lockfile for relaycast sdk 1.0.0 bump
* fix: bump gemini relay extension to relaycast mcp 1.0.0
* chore(release): v3.2.9
* feat(sdk): polish workflow CLI output with listr2 spinners and chalk colors
- Replace plain console.log progress in cli.ts with listr2 task list
- Per-step spinners show owner, retry, nudge, force-release, and review events
- chalk colors: cyan for timestamps, green/red/yellow for status, dim for metadata
- logRunSummary() and broker stderr use chalk for visual hierarchy
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: MCP tools unavailable for agents spawned via agent_add
Two root causes prevented agents spawned via the Relaycast API
(agent_add MCP tool) from loading MCP tools:
1. Claude: --strict-mcp-config blocked .mcp.json loading. Removed it
so --mcp-config is additive — only passes relaycast config while
Claude loads user MCP servers from .mcp.json independently.
2. All CLIs: The WS AgentSpawnRequested handler had no agent
pre-registration. The AgentSpawnRequestedPayload struct doesn't
include a token field, so relaycast_ws_spawn_token() always
returned None. Added broker-side register_agent_token() calls
(matching the SDK spawn_agent path) to both WS spawn handlers.
Tests:
- New integration test (agent-spawns-agent.test.ts) exercises the
exact agent_add flow for claude, codex, and gemini
- Updated unit tests and e2e tests for new --mcp-config behavior
- All 220 lib + 8 e2e Rust tests pass
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* style: auto-format Rust code with cargo fmt
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: GitHub Actions <actions@github.com>
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
khaliqgant
added a commit
that referenced
this pull request
Mar 19, 2026
* feat: add workflow to polish CLI output with listr2 + chalk
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* ci: add workflow validation and dry-run check on PR
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(ci): fetch full history for git diff, run dry-run on all workflow types
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(sdk): polish workflow CLI output with listr2 spinners and chalk colors
- Replace plain console.log progress in cli.ts with listr2 task list
- Per-step spinners show owner, retry, nudge, force-release, and review events
- chalk colors: cyan for timestamps, green/red/yellow for status, dim for metadata
- logRunSummary() and broker stderr use chalk for visual hierarchy
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* test: add smoke test workflow for listr2 output rendering
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: use ESM-compatible import for check-sdk step
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(sdk): export createWorkflowRenderer for listr2 output in TS workflows
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: pre-attach catch to prevent unhandled rejection on fast-failing steps
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: use worker preset for verify step, mute console during listr rendering
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: filter [broker]/[workflow] noise, show observer URL, add unmount()
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: add missing unhandled-rejection guards and output filter to YAML path
- cli.ts: installOutputFilter() in runWithListr so YAML workflows also
suppress [broker]/[workflow HH:MM] noise during listr rendering
- cli.ts: done.catch()/workflowDone.catch() guards for fast-failing steps
- listr-renderer.ts: workflowDone.catch() guard for instant run:failed
- listr-renderer.ts: add renderer.unmount() to JSDoc example
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: address PR review feedback for workflow output polish
- Output filter: use regex .test() instead of startsWith/^ anchor so
chalk-colored [broker] and [workflow HH:MM] lines are properly suppressed
- Resume mode: add event listener for step progress reporting
- GHA workflow: fix grep exit code with || true, use env var instead of
raw ${{ }} interpolation (script injection), use npx tsx instead of
non-existent 'run' subcommand, only validate/dry-run YAML files
- Workflow: fix incorrect CJS assumption (SDK is ESM), add final
type-check gate after review step
* fix: CI failures and skipped-step visibility
- Add chalk and listr2 to root package.json (Build & Validate requires them)
- Dynamic import listr2 so SDK loads on Node 18 (styleText not available)
- Show steps skipped without prior start event in listr output
- Remove unused ListrType import
* fix: detect claude CLI with inline args for MCP injection (#584)
* fix: detect claude CLI with inline args for MCP injection
* fix: extract executable from cli string in gemini/droid mcp setup
When cli contains inline args (e.g. 'gemini --model foo'),
Command::new(cli) fails because it treats the entire string as
an executable path. Now extract just the binary via shlex::split
before passing to Command::new and manual_cmd.
* chore(release): v3.2.8
* bump versions (#590)
* bump versions
* fix: refresh lockfile for relaycast sdk 1.0.0 bump
* fix: bump gemini relay extension to relaycast mcp 1.0.0
* chore(release): v3.2.9
* feat(sdk): polish workflow CLI output with listr2 spinners and chalk colors
- Replace plain console.log progress in cli.ts with listr2 task list
- Per-step spinners show owner, retry, nudge, force-release, and review events
- chalk colors: cyan for timestamps, green/red/yellow for status, dim for metadata
- logRunSummary() and broker stderr use chalk for visual hierarchy
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: regenerate lockfile with npm 11 for Node 24 ci compatibility
* fix: address remaining workflow review feedback
- pass yaml path before --validate in workflow validation CI
- add trail start/complete/abandon to polish workflow script
- fix remaining ESM wording/install command in planning prompt
- queue listr tasks before lazy renderer initialization completes
* fix: mark test-only MCP merge helpers as dead-code allowed
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: GitHub Actions <actions@github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Why
PR #583 fixed workspace context forwarding, but spawned agents can still miss Relaycast MCP tools when the configured CLI string includes inline args. In that case provider detection could fail and skip MCP config injection.
Testing