feat(workflow): expose dynamic workflows through MCP by Waishnav · Pull Request #98 · Waishnav/devspace

Waishnav · 2026-07-21T16:24:29Z

Dynamic Workflows were reachable only from the local CLI, so an MCP client could not start or follow durable runs without dropping down to shell commands. The model also lacked focused guidance for choosing workflows and available local-agent providers.

Expose run, status, and cancellation through MCP while keeping execution in the detached worker, so workflows continue after an MCP request disconnects. Add the Dynamic Workflows skill, seed the bundled agent skills consistently, and introduce runtime provider availability to guide workflow execution.

Stack

Based on #97 and merged into #94 before provider-native structured output in #99.

Always include bundled skills root when subagents enabled so seeding subagent-delegation no longer hides later skills. Seed dynamic-workflows alongside subagent-delegation on init. Co-Authored-By: Claude <noreply@anthropic.com>

Load ordered enable-list from config/env, probe available providers on init and doctor, and filter workflow CLI providers by enabled∩live. Co-Authored-By: Claude <noreply@anthropic.com>

Gate on config.subagents. run_workflow spawns the same detached worker as CLI; status long-polls journal events; cancel requests cooperative stop. Co-Authored-By: Claude <noreply@anthropic.com>

coderabbitai · 2026-07-21T16:24:35Z

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 38401585-7d31-43df-a24e-6165b003ebcd

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests
Commit unit tests in branch pr/dw-5-mcp-skill

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

Waishnav · 2026-07-21T17:25:36Z

+      title: "Workflow status",
+      description: "Drain events for a workflow run; optional long-poll yield.",
+      inputSchema: {
+        runId: z.string(),


[P1] Status is authorized only by a global runId and is not scoped back to the open workspace. Please include workspaceId (or enforce equivalent ownership) and verify the run's workspaceId/root before returning events. The same check is needed for cancel and resume.

Intentionally not adopting mandatory workspace-scoped status/cancel authorization. DevSpace workflows are also a portable dynamic-workflow layer for Cursor, Pi, OpenCode, and other harnesses that may invoke CLI/MCP without owning a DevSpace workspaceId. runId remains the durable cross-harness handle by product decision. Filesystem confinement is handled separately through project-local workflow resolution rather than making workspace IDs mandatory.

Waishnav · 2026-07-21T17:25:38Z

+        const run = store.requestCancel(runId);
+        if (run.pid && (run.status === "running" || run.status === "starting")) {
+          try {
+            process.kill(run.pid, "SIGTERM");


[P0] MCP cancellation has no grace-period supervision or hard-kill/terminal fallback. It sends one SIGTERM and can leave the database row running forever if the worker dies before journaling. Please reuse the same shared cancellation supervisor as the CLI and guarantee run_cancelled plus terminal status.

Addressed by follow-up PR #115. MCP and CLI now call the same cancelWorkflowRun() supervisor, including cooperative grace, cross-platform process-tree termination, hard-kill escalation, and an atomic run_cancelled terminal fallback. MCP cancellation no longer returns after a single best-effort SIGTERM.

Waishnav · 2026-07-21T17:25:42Z

agentProviders.enabled is currently applied inside workflow execution, but the rest of the agent surface is inconsistent: open_workspace still advertises every live provider and one-shot agent paths can use providers that the user disabled. Please centralize provider policy and reuse it for workflows, agents run, open_workspace, and doctor.

This PR also needs integration coverage for MCP disconnect survival, cancellation reaching a terminal state, paginated terminal event draining, and provider-config filtering.

Wire LocalAgentRunInput.schema through CodexSdkLocalAgentRuntime to thread.run turn options, and surface parsed structured output.

Pass JSON Schema via outputFormat on query options and prefer structured_output from result messages. OpenCode stays prompt-path only.

Hardcode NATIVE_SCHEMA_PROVIDERS; attempt 0 uses adapter schema without prompt bloat, then prompt-repair retries with Ajv. Wire schema through runProvider / CLI worker. Document in skill.

refactor(agents): unify provider resolution and subagent guidance

feat(workflow): support agent profiles and project workflows

fix(workflow): replay deterministic call prefixes

fix(workflow): supervise cancellation and terminal lifecycle

feat(workflow): isolate workflow script execution

feat(ui): live workspace and workflow dashboards

feat(workflow): read-only project workflow TUI

feat(workflow): script iteration, replay provenance, and call inspection

refactor(workflow): model operational failures with better-result

refactor(workflow): strengthen contracts and zod boundaries

feat(agents): native schema for codex/claude agent({ schema })

…veable in Docker / script mode (Waishnav#100) (Waishnav#102) * fix(frontend): survive browser page translation instead of crashing the whole tree (Waishnav#100) Chrome / Edge 'Translate this page' replaces text nodes with <font> wrappers; React's next removeChild / insertBefore on those nodes throws NotFoundError and the ErrorBoundary unmounts everything. Guard both DOM methods before mount (per react/react#11538) so displaced nodes are skipped rather than fatal. ErrorBoundary now detects the translation case and tells the user (zh/en) to turn translation off; its fallback page is redone per DESIGN.md (no purple gradient / AntD Result, single-tone card + Btn primitives) and 'back home' actually navigates under HashRouter. Co-authored-by: Kris K <zhouxiaoka@users.noreply.github.com> * feat(settings): make the settings page work in Docker / script mode (Waishnav#100) GET/PUT /settings, /test-api, /current-provider, /compatible-models, /available-models, /validation, /reset and DELETE / no longer require desktop mode. settings.json lands in get_data_directory() (./data in Docker) and the API process + Celery worker already hot-reload on mtime, so the same code path works for every deployment shape. Desktop-only endpoints (data-dir migration, backup/restore, import/export, config sync) keep the guard. When no settings.json exists yet, GET /settings reflects the provider / model / base_url the LLM manager actually resolved from env (LLM_PROVIDER etc.) instead of always showing dashscope. Frontend SettingsPage drops the web-mode read-only short-circuit; also removes the unused eslint-disable directive that has kept CI red on main (Waishnav#99). Adds backend/tests/test_settings_web_mode.py (6 cases). Co-authored-by: Kris K <zhouxiaoka@users.noreply.github.com> * docs: record Waishnav#100 fix, 2026-09-20 issue/PR triage and schedule feature requests into v1.3 / v1.4 HANDOFF: root-cause note for Waishnav#100, per-issue disposition table, external PR decisions (Waishnav#97 Waishnav#98 Waishnav#99 merge, Waishnav#101 hold for i18n scaffold), and new v1.3 items (fail-like-failure, i18n scaffold, dashscope intl, DESIGN.md debt list) plus v1.4 items (pluggable ASR / scoring backends, Sentry). DOCKER.md: settings can now be saved from the web UI; env vars are defaults only. Co-authored-by: Kris K <zhouxiaoka@users.noreply.github.com> --------- Co-authored-by: Cursor Agent <cursoragent@cursor.com> Co-authored-by: Kris K <zhouxiaoka@users.noreply.github.com>

…er-queues

Waishnav and others added 3 commits July 21, 2026 17:42

feat(agents): add agentProviders config + init/doctor probe

349b52c

Load ordered enable-list from config/env, probe available providers on init and doctor, and filter workflow CLI providers by enabled∩live. Co-Authored-By: Claude <noreply@anthropic.com>

feat(workflow): register MCP run/status/cancel tools

e1ec049

Gate on config.subagents. run_workflow spawns the same detached worker as CLI; status long-polls journal events; cancel requests cooperative stop. Co-Authored-By: Claude <noreply@anthropic.com>

Waishnav commented Jul 21, 2026

View reviewed changes

Waishnav added 3 commits July 21, 2026 22:55

feat(agents): pass schema as Codex turn outputSchema

3030d86

Wire LocalAgentRunInput.schema through CodexSdkLocalAgentRuntime to thread.run turn options, and surface parsed structured output.

feat(agents): Claude native outputFormat + structured_output

6af99da

Pass JSON Schema via outputFormat on query options and prefer structured_output from result messages. OpenCode stays prompt-path only.

feat(workflow): native-first schema for codex/claude agent()

7d797e3

Hardcode NATIVE_SCHEMA_PROVIDERS; attempt 0 uses adapter schema without prompt bloat, then prompt-repair retries with Ajv. Wire schema through runProvider / CLI worker. Document in skill.

Waishnav mentioned this pull request Jul 21, 2026

feat(workflow): use native structured output for Codex and Claude #99

Merged

Waishnav and others added 15 commits July 22, 2026 00:22

fix(workflow): make native schema fallback session-aware

a8b8e5a

test(workflow): normalize persisted paths on windows

3343404

refactor(workflow): strengthen contracts and zod boundaries

4cb4007

refactor(workflow): model operational failures with better-result

d5b00f5

feat(workflow): persist replay provenance

f9a1dc7

feat(workflow): add script iteration and call inspection

c8f0b65

docs(workflow): teach recovery without budget stub

058297b

feat(workflow): query runs by workspace

332aa8e

feat(workflow): project workflow read model

a499c2e

feat(workflow): add read-only project TUI

5ee0022

feat(workflow): add UI snapshot projections

dbe26f7

feat(workflow): expose read-only app snapshots

d3643e8

feat(ui): recognize workflow result cards

57296a6

feat(ui): add live workspace and workflow dashboards

4bc9558

docs(ui): document read-only workflow dashboards

4777bab

coderabbitai Bot approved these changes Jul 26, 2026

View reviewed changes

Waishnav added 3 commits July 26, 2026 20:06

feat(workflow): isolate script execution in child process

f5e9252

test(workflow): cover sandbox termination and escapes

09a7fc4

fix(workflow): keep sandbox values in vm realm

12cee07

Waishnav added 24 commits July 27, 2026 13:01

refactor(config): remove experimental provider persistence

620b180

refactor(agents): share execution target resolution

d640cf8

refactor(agents): describe provider model and effort capabilities

f2482bb

refactor(features): separate subagent and workflow capabilities

321b81c

fix(workspace): expose only usable agent capabilities

69fee30

refactor(skills): replace subagent delegation guidance

f25d8e9

docs(skills): add provider-specific override references

2f17478

fix(skills): keep bundled agent skills package-managed

f214c19

refactor(workflow): name live provider availability accurately

1ace1f0

docs(agents): document experimental capability boundaries

5c1ddea

test(workflow): follow provider availability rename

691998e

docs(config): describe effective agent capability gates

5bf5374

feat(agents): discover usable delegation targets

562e54a

feat(workflow): merge dw-16 agent/subagent overhaul (#118)

b347a82

refactor(agents): unify provider resolution and subagent guidance

feat(workflow): merge dw-15 profiles and project workflows (#117)

b677c3e

feat(workflow): support agent profiles and project workflows

Merge pull request #116 from pr/dw-14-prefix-replay-integrity

8ce8274

fix(workflow): replay deterministic call prefixes

Merge pull request #115 from pr/dw-13-workflow-lifecycle-supervisor

0b0f5e2

fix(workflow): supervise cancellation and terminal lifecycle

Merge pull request #114 from pr/dw-12-sandbox-process-boundary

e3ffd22

feat(workflow): isolate workflow script execution

Merge pull request #112 from pr/dw-11-workflow-mcp-ui

0328090

feat(ui): live workspace and workflow dashboards

Merge pull request #111 from pr/dw-10-workflow-tui

6b449e6

feat(workflow): read-only project workflow TUI

Merge pull request #108 from pr/dw-9-resume-iteration-inspection

1a6764d

feat(workflow): script iteration, replay provenance, and call inspection

Merge pull request #101 from pr/dw-8-result-errors

1e21ad1

refactor(workflow): model operational failures with better-result

Merge pull request #100 from pr/dw-7-typed-contracts-zod

d052520

refactor(workflow): strengthen contracts and zod boundaries

Merge pull request #99 from pr/dw-6-native-schema

7304a6a

feat(agents): native schema for codex/claude agent({ schema })

Waishnav marked this pull request as ready for review July 27, 2026 15:05

Waishnav merged commit 171d43f into pr/dw-4-cli-worker Jul 27, 2026
1 check was pending

Waishnav mentioned this pull request Jul 30, 2026

feat(workflow): run and resume workflows from the CLI #97

Merged

Waishnav changed the title ~~feat(workflow): MCP tools, skill, agentProviders~~ feat(workflow): expose dynamic workflows through MCP Jul 30, 2026

SoloistHart pushed a commit to SoloistHart/autoclip that referenced this pull request Sep 27, 2026

Merge pull request Waishnav#98 from mitsa-ng/feature/docs-celery-work…

9796e86

…er-queues

Navigation Menu

Repository navigation

feat(workflow): expose dynamic workflows through MCP - #98

feat(workflow): expose dynamic workflows through MCP#98
Waishnav merged 67 commits into
pr/dw-4-cli-workerfrom
pr/dw-5-mcp-skill

Waishnav commented Jul 21, 2026 •

edited

Loading

Uh oh!

coderabbitai Bot commented Jul 21, 2026 •

edited

Loading

Review skipped

Uh oh!

Waishnav Jul 21, 2026

Uh oh!

Waishnav Jul 26, 2026

Uh oh!

Waishnav Jul 21, 2026

Uh oh!

Waishnav Jul 26, 2026

Uh oh!

Waishnav commented Jul 21, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

Waishnav commented Jul 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Stack

Uh oh!

coderabbitai Bot commented Jul 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review skipped

Uh oh!

Waishnav Jul 21, 2026

Choose a reason for hiding this comment

Uh oh!

Waishnav Jul 26, 2026

Choose a reason for hiding this comment

Uh oh!

Waishnav Jul 21, 2026

Choose a reason for hiding this comment

Uh oh!

Waishnav Jul 26, 2026

Choose a reason for hiding this comment

Uh oh!

Waishnav commented Jul 21, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Waishnav commented Jul 21, 2026 •

edited

Loading

coderabbitai Bot commented Jul 21, 2026 •

edited

Loading