Skip to content

armgit / mso

Public Imported from GitHub
0

Browser-based graphical shell and control plane for a Linux server you own — real terminal, file manager, live metrics and BYOK AI in one mobile-first workspace. Next.js 16 · React 19 · node-pty · Tailwind v4 · shadcn/ui. Public alpha, self-hosted.

fix(workflows): compact graph mcp surface

Rahmanfc22f855e09c
diff --git a/docs/CHANGELOG.md b/docs/CHANGELOG.md
index 97dd0597..e148c6af 100644
--- a/docs/CHANGELOG.md
+++ b/docs/CHANGELOG.md
@@ -12,6 +12,7 @@ this is the *what*, and it is what Settings → About shows as “What's new”.
**Fixed**
+- `workflows` compact graph mcp surface
- `workflows` complete cli and release gates
- `mcp` expose downstream project tool results (#43)
diff --git a/docs/WORKFLOW-GRAPH.md b/docs/WORKFLOW-GRAPH.md
index 87c408ba..67f8618f 100644
--- a/docs/WORKFLOW-GRAPH.md
+++ b/docs/WORKFLOW-GRAPH.md
@@ -78,10 +78,7 @@ This makes repeated session history useful without requiring a separate manual m
## MCP tools
-- `workflow_graph_catalog`
-- `workflow_graph_manage`
-- `workflow_graph_run`
-- `workflow_graph_status`
+- `workflow_graph` — `action=list|get|create|update|delete|clone|run|status`; action-specific fields go in `data` (`definition`, `revision`, `input`, `key`).
The existing v1 tools remain compatible:
diff --git a/docs/generated/MCP-CATALOG.md b/docs/generated/MCP-CATALOG.md
index 69d52d70..7eccb157 100644
--- a/docs/generated/MCP-CATALOG.md
+++ b/docs/generated/MCP-CATALOG.md
@@ -4,21 +4,21 @@
## Full MSO catalog
-<!-- mcp-toolset: server=1.14.4 version=2026.09.15.3 tools=109 read=52 write=35 exec=22 -->
+<!-- mcp-toolset: server=1.14.4 version=2026.09.15.3 tools=106 read=50 write=34 exec=22 -->
| Fact | Current source value |
|---|---:|
| MCP server | `1.14.4` |
| Toolset | `2026.09.15.3` |
| Toolset changed at | `2026-09-15T03:35:00+07:00` |
-| Transport tools | **111** |
-| Model/operator tools | **109** |
-| Read | **52** |
-| Write | **35** |
+| Transport tools | **108** |
+| Model/operator tools | **106** |
+| Read | **50** |
+| Write | **34** |
| Exec | **22** |
| App-only bridges | **2** |
-### Read (52)
+### Read (50)
- `a2a_agent_discover`
- `a2a_agents_list`
@@ -70,10 +70,8 @@
- `sys_stats`
- `tool_forge_candidates`
- `vps_status`
-- `workflow_graph_catalog`
-- `workflow_graph_status`
-### Write (35)
+### Write (34)
- `a2a_agent_register`
- `a2a_agent_remove`
@@ -108,7 +106,6 @@
- `tool_forge_propose`
- `workflow_cancel`
- `workflow_finish`
-- `workflow_graph_manage`
- `workflow_start`
### Exec (22)
@@ -134,7 +131,7 @@
- `project_mcp_tools`
- `tool_forge_evaluate`
- `tool_forge_promote`
-- `workflow_graph_run`
+- `workflow_graph`
### App-only bridges (2)
@@ -143,20 +140,20 @@
## ChatGPT model profile
-<!-- mcp-chatgpt-profile: server=1.14.4 version=2026.09.15.3 tools=109 read=52 write=35 exec=22 app-only=2 total=111 -->
+<!-- mcp-chatgpt-profile: server=1.14.4 version=2026.09.15.3 tools=106 read=50 write=34 exec=22 app-only=2 total=108 -->
The ChatGPT profile automatically projects the complete MSO-owned generic model/operator catalog. OAuth scope is still enforced independently; app-only compatibility bridges stay app-only, and project-owned MCP tool names remain dynamic data behind the generic project bridge.
| Fact | Current source value |
|---|---:|
-| ChatGPT transport tools | **111** |
-| ChatGPT model/operator tools | **109** |
-| Read | **52** |
-| Write | **35** |
+| ChatGPT transport tools | **108** |
+| ChatGPT model/operator tools | **106** |
+| Read | **50** |
+| Write | **34** |
| Exec | **22** |
| App-only bridges | **2** |
-### ChatGPT read (52)
+### ChatGPT read (50)
- `a2a_agent_discover`
- `a2a_agents_list`
@@ -208,10 +205,8 @@ The ChatGPT profile automatically projects the complete MSO-owned generic model/
- `sys_stats`
- `tool_forge_candidates`
- `vps_status`
-- `workflow_graph_catalog`
-- `workflow_graph_status`
-### ChatGPT write (35)
+### ChatGPT write (34)
- `a2a_agent_register`
- `a2a_agent_remove`
@@ -246,7 +241,6 @@ The ChatGPT profile automatically projects the complete MSO-owned generic model/
- `tool_forge_propose`
- `workflow_cancel`
- `workflow_finish`
-- `workflow_graph_manage`
- `workflow_start`
### ChatGPT exec (22)
@@ -272,7 +266,7 @@ The ChatGPT profile automatically projects the complete MSO-owned generic model/
- `project_mcp_tools`
- `tool_forge_evaluate`
- `tool_forge_promote`
-- `workflow_graph_run`
+- `workflow_graph`
### ChatGPT app-only bridges (2)
diff --git a/lib/mcp/parity.test.ts b/lib/mcp/parity.test.ts
index acc895ff..68d2aaf9 100644
--- a/lib/mcp/parity.test.ts
+++ b/lib/mcp/parity.test.ts
@@ -21,10 +21,7 @@ const ALFA_ONLY: Record<string, string> = {
};
const MCP_ONLY: Record<string, string> = {
- "workflow.graph.catalog": "Native Workflows UI uses the private REST surface; MCP agents need a principal-scoped catalog without inheriting an unrelated Alfa browser-thread identity.",
- "workflow.graph.manage": "Workflow graph CRUD is exposed natively by the Workflows app and separately to MCP principals; sharing Alfa thread identity would collapse the deliberate principal boundary.",
- "workflow.graph.run": "Server-native graph execution is principal/idempotency scoped; Alfa uses the native Workflows app rather than claiming an external MCP workflow principal.",
- "workflow.graph.status": "Per-node graph receipts belong to the exact workflow principal; the native Workflows app has its own authenticated receipt surface.",
+ "workflow.graph": "Native Workflows UI has its own authenticated principal; external MCP graph CRUD/run/status must stay principal-scoped instead of inheriting an unrelated browser identity.",
"flow.catalog": "Project API/MCP automation belongs to the external/CLI session context; Alfa has its own browser workflow identity.",
"flow.run": "Provider mutations require the durable external principal, exact connection, and idempotency receipt; owner CLI shares this executor.",
"flow.status": "Flow receipts belong to their authenticated external/CLI principal; Alfa must not inherit that principal.",
@@ -169,7 +166,7 @@ describe("MCP rate limits mirror the routes", () => {
"fs.delete": 60, "fs.upload": 20, "exec": 60, "managed-app": 12, "camoufox": 12,
};
const MCP_NATIVE_LIMITS: Record<string, number> = {
- "flow.run": 20, "flow.manage": 20, "workflow.graph.manage": 30, "workflow.graph.run": 20, "agent.session.open": 20, "project.mcp.manage": 20, "project.asset.attach": 20,
+ "flow.run": 20, "flow.manage": 20, "workflow.graph": 20, "agent.session.open": 20, "project.mcp.manage": 20, "project.asset.attach": 20,
// Native capabilities have independent bounded buckets (no matching HTTP route).
"screen.capture": 10,
"fs.export": 20,
diff --git a/lib/mcp/tools-flows.ts b/lib/mcp/tools-flows.ts
index ad741d9a..9494a816 100644
--- a/lib/mcp/tools-flows.ts
+++ b/lib/mcp/tools-flows.ts
@@ -5,8 +5,8 @@ import { manageProjectFlow } from "@/lib/host/project-flow-manifest";
import { type McpTool, S, str } from "./tool-kit";
import { cloneWorkflowGraph, createWorkflowGraph, deleteWorkflowGraph, getWorkflowGraph, listWorkflowGraphs, updateWorkflowGraph } from "@/lib/workflow/graph-store";
import { startWorkflowGraph, workflowGraphRunStatus } from "@/lib/workflow/graph-engine";
-const project = { type: "string", maxLength: 4096, description: "Exact project id/path/name." };
-const flow = { type: "string", maxLength: 64, description: "Flow id from flow_catalog." };
+const project = { type: "string", maxLength: 4096 };
+const flow = { type: "string", maxLength: 64 };
function privateGraphPrincipal(context: Parameters<McpTool["run"]>[1]): string {
const principal = context.recipeActor ?? context.workflowActor ?? context.actor ?? context.principal;
@@ -15,14 +15,14 @@ function privateGraphPrincipal(context: Parameters<McpTool["run"]>[1]): string {
}
export const FLOW_TOOLS: McpTool[] = [
{ name: "flow_catalog", title: "Discover Automation Flows", scope: "read",
- description: "Discover reusable API/MCP flows for an exact project. Returns input schema, ordered steps, account context and required scope. Inspect a flow before flow_run; project flows live in .mso/flows.json.",
- chatgptDescription: "Discover project automation flows, required inputs and ordered steps. Inspect before flow_run.",
+ description: "Discover project API/MCP flows, schemas and ordered steps. Inspect before flow_run.",
+ chatgptDescription: "Discover project flows and inputs before flow_run.",
annotations: { readOnlyHint: true, destructiveHint: false, openWorldHint: false },
inputSchema: S({ project, flow }, ["project"]),
run: async a => { const { project: p, flows, revision } = await flowCatalog(str(a, "project"), typeof a.flow === "string" ? a.flow : undefined); return { project: p.id, revision, flows }; } },
{ name: "flow_run", title: "Run Automation Flow", scope: "exec",
- description: "Start one inspected API/MCP flow with exact project, input and idempotency_key. Child tools retain scope, connection guards and audits. Same key/input returns the existing receipt; changed input is refused. Returns durable run id; use flow_status for progress. Never auto-retries a failed mutation.",
- chatgptDescription: "Run an inspected flow for exact project/input/idempotency_key. Returns a run id; flow_status reads progress.",
+ description: "Run an inspected project flow with exact input/idempotency key. Returns a durable receipt; failed writes are not auto-retried.",
+ chatgptDescription: "Run an inspected flow; use flow_status for its receipt.",
annotations: { readOnlyHint: false, destructiveHint: true, openWorldHint: true }, audit: { action: "exec.run", targetArg: "flow" },
limit: { key: "flow.run", max: 20, windowMs: 60_000 },
inputSchema: S({ project, flow, input: { type: "object", additionalProperties: true }, idempotency_key: { type: "string", minLength: 1, maxLength: 128 } }, ["project", "flow", "input", "idempotency_key"]),
@@ -32,14 +32,14 @@ export const FLOW_TOOLS: McpTool[] = [
return startFlow(catalog.definitions[0], catalog.project.id, a.input, str(a, "idempotency_key"), context, name => TOOLS_BY_NAME.get(name));
} },
{ name: "flow_status", title: "Automation Flow Status", scope: "read",
- description: "Read a durable flow receipt owned by this authenticated principal. Optional wait_ms (0-25000) reduces client polling. Reports step results and interrupted outcomes; never replays work.",
- chatgptDescription: "Read a flow run's step results. wait_ms up to 25000 reduces polling. Interrupted writes need outcome inspection.",
+ description: "Read this principal's flow receipt; wait_ms may reduce polling.",
+ chatgptDescription: "Read flow step results; wait_ms may reduce polling.",
annotations: { readOnlyHint: true, destructiveHint: false, openWorldHint: false },
inputSchema: S({ run_id: { type: "string", pattern: "^[a-f0-9]{32}$" }, wait_ms: { type: "integer", minimum: 0, maximum: 25_000 } }, ["run_id"]),
run: (a, context) => flowStatus(str(a, "run_id"), context, Number(a.wait_ms) || 0) },
{ name: "flow_manage", limit: { key: "flow.manage", max: 20, windowMs: 60_000 }, title: "Save Project Automation", scope: "write",
- description: "Create/update/delete one project flow in .mso/flows.json using the revision from flow_catalog (new for absent manifest). Built-ins cannot be overwritten. Definitions contain metadata and exact API/MCP tool steps; credentials are forbidden.",
- chatgptDescription: "Save or delete a project flow with revision from flow_catalog. Never include credentials; built-ins are immutable.",
+ description: "Create/update/delete a project flow with flow_catalog revision. Built-ins are immutable; credentials are forbidden.",
+ chatgptDescription: "Save/delete a project flow by revision; no credentials.",
annotations: { readOnlyHint: false, destructiveHint: true, openWorldHint: false }, audit: { action: "fs.write", targetArg: "project" },
inputSchema: S({ project, flow, action: { type: "string", enum: ["upsert", "delete"] }, revision: { type: "string" }, definition: { type: "object", additionalProperties: true } }, ["project", "flow", "action", "revision"]),
run: async a => {
@@ -49,46 +49,27 @@ export const FLOW_TOOLS: McpTool[] = [
const catalog = await flowCatalog(str(a, "project"));
return manageProjectFlow(catalog.project.path, { action: a.action, id, flow: a.definition, revision: str(a, "revision") });
} },
- { name: "workflow_graph_catalog", title: "Workflow Graph Catalog", scope: "read",
- description: "List private Workflow Graph v2 definitions for this authenticated principal, or inspect one exact graph. Graphs are isolated per principal and never contain provider secrets.",
- chatgptDescription: "List or inspect this user's private n8n-like workflow graphs before managing/running one.",
- annotations: { readOnlyHint: true, destructiveHint: false, openWorldHint: false },
- inputSchema: S({ graph_id: { type: "string", maxLength: 96 } }),
- run: async (a, context) => {
- const principal = privateGraphPrincipal(context);
- if (typeof a.graph_id === "string" && a.graph_id) { const graph = await getWorkflowGraph(principal, a.graph_id); if (!graph) throw new Error("workflow graph not found"); return { graph }; }
- const graphs = await listWorkflowGraphs(principal);
- return { graphs: graphs.map(graph => ({ id: graph.id, name: graph.name, description: graph.description, status: graph.status, revision: graph.revision, nodeCount: graph.nodes.length, updatedAt: graph.updatedAt, provenance: graph.metadata.provenance, project: graph.metadata.project })) };
- } },
- { name: "workflow_graph_manage", title: "Manage Workflow Graph", scope: "write", limit: { key: "workflow.graph.manage", max: 30, windowMs: 60_000 },
- description: "Create, update, clone or delete one private Workflow Graph v2 definition. Updates/deletes require expected_revision. Definitions are metadata-only; secret-like fields are rejected.",
- chatgptDescription: "CRUD a private n8n-like workflow graph with revision conflict protection.",
- annotations: { readOnlyHint: false, destructiveHint: true, openWorldHint: false }, audit: { action: "fs.write", targetArg: "graph_id" },
- inputSchema: S({ action: { type: "string", enum: ["create", "update", "delete", "clone"] }, graph_id: { type: "string", maxLength: 96 }, expected_revision: { type: "string", maxLength: 128 }, definition: { type: "object", additionalProperties: true } }, ["action"]),
+ { name: "workflow_graph", title: "Workflow Graph", scope: "exec", limit: { key: "workflow.graph", max: 20, windowMs: 60_000 },
+ description: "Private Workflow Graph v2 CRUD/run/status. Use action list|get|create|update|delete|clone|run|status. Definitions reject secrets; runs keep per-node receipts.",
+ chatgptDescription: "CRUD/run workflows; read node logs.",
+ annotations: { readOnlyHint: false, destructiveHint: true, openWorldHint: true }, audit: { action: "exec.run", targetArg: "id" },
+ inputSchema: S({ action: { type: "string", enum: ["list", "get", "create", "update", "delete", "clone", "run", "status"] }, id: { type: "string", maxLength: 96 }, data: { type: "object", additionalProperties: true }, wait_ms: { type: "integer", minimum: 0, maximum: 25_000 } }, ["action"]),
run: async (a, context) => {
const principal = privateGraphPrincipal(context), action = str(a, "action");
- if (action === "create") return { graph: await createWorkflowGraph(principal, a.definition) };
- const id = str(a, "graph_id");
+ const data = a.data && typeof a.data === "object" && !Array.isArray(a.data) ? a.data as Record<string, unknown> : {};
+ if (action === "list") { const graphs = await listWorkflowGraphs(principal); return { graphs: graphs.map(graph => ({ id: graph.id, name: graph.name, status: graph.status, revision: graph.revision, nodeCount: graph.nodes.length, updatedAt: graph.updatedAt, provenance: graph.metadata.provenance, project: graph.metadata.project })) }; }
+ if (action === "status") return workflowGraphRunStatus(principal, str(a, "id"), Number(a.wait_ms) || 0);
+ if (action === "create") return { graph: await createWorkflowGraph(principal, data.definition ?? data) };
+ const id = str(a, "id");
+ if (action === "get") { const graph = await getWorkflowGraph(principal, id); if (!graph) throw new Error("workflow graph not found"); return { graph }; }
if (action === "clone") return { graph: await cloneWorkflowGraph(principal, id) };
- if (action === "update") return { graph: await updateWorkflowGraph(principal, id, str(a, "expected_revision"), a.definition) };
- if (action === "delete") return deleteWorkflowGraph(principal, id, str(a, "expected_revision"));
+ if (action === "update") return { graph: await updateWorkflowGraph(principal, id, String(data.revision ?? ""), data.definition) };
+ if (action === "delete") return deleteWorkflowGraph(principal, id, String(data.revision ?? ""));
+ if (action === "run") {
+ const graph = await getWorkflowGraph(principal, id); if (!graph) throw new Error("workflow graph not found");
+ const { TOOLS_BY_NAME } = await import("./tools");
+ return startWorkflowGraph(graph, data.input ?? {}, String(data.key ?? ""), context, name => TOOLS_BY_NAME.get(name), principal);
+ }
throw new Error("unsupported workflow graph action");
} },
- { name: "workflow_graph_run", title: "Run Workflow Graph", scope: "exec", limit: { key: "workflow.graph.run", max: 20, windowMs: 60_000 },
- description: "Run one inspected private Workflow Graph v2 with exact input/idempotency key. Execution remains bounded by normal capability scopes and returns a durable per-node receipt.",
- chatgptDescription: "Run a private workflow graph and return the per-node run receipt id.",
- annotations: { readOnlyHint: false, destructiveHint: true, openWorldHint: true }, audit: { action: "exec.run", targetArg: "graph_id" },
- inputSchema: S({ graph_id: { type: "string", maxLength: 96 }, input: { type: "object", additionalProperties: true }, idempotency_key: { type: "string", minLength: 1, maxLength: 128 } }, ["graph_id", "input", "idempotency_key"]),
- run: async (a, context) => {
- const principal = privateGraphPrincipal(context), graph = await getWorkflowGraph(principal, str(a, "graph_id"));
- if (!graph) throw new Error("workflow graph not found");
- const { TOOLS_BY_NAME } = await import("./tools");
- return startWorkflowGraph(graph, a.input, str(a, "idempotency_key"), context, name => TOOLS_BY_NAME.get(name), principal);
- } },
- { name: "workflow_graph_status", title: "Workflow Graph Run Status", scope: "read",
- description: "Read a private Workflow Graph v2 run receipt with node states, logs, durations and failed-node identity. Optional wait_ms reduces polling.",
- chatgptDescription: "Read node-by-node workflow graph status/logs and identify the failed node.",
- annotations: { readOnlyHint: true, destructiveHint: false, openWorldHint: false },
- inputSchema: S({ run_id: { type: "string", pattern: "^[a-f0-9]{32}$" }, wait_ms: { type: "integer", minimum: 0, maximum: 25_000 } }, ["run_id"]),
- run: (a, context) => workflowGraphRunStatus(privateGraphPrincipal(context), str(a, "run_id"), Number(a.wait_ms) || 0) },
];
diff --git a/lib/mcp/tools-workflow-lifecycle.ts b/lib/mcp/tools-workflow-lifecycle.ts
index 72f3956c..7ee660e3 100644
--- a/lib/mcp/tools-workflow-lifecycle.ts
+++ b/lib/mcp/tools-workflow-lifecycle.ts
@@ -12,8 +12,7 @@ import { evidenceInput, evidenceSchema, WORKFLOW_PROGRESS_OUTPUT, workflowProgre
export const WORKFLOW_LIFECYCLE_TOOLS: McpTool[] = [
{
name: "workflow_status",
- description:
- "Return a redacted live status snapshot for one workflow. This remains app-only for compatibility with cached progress widgets and MSO operator views; workflow_start itself is headless. It exposes only workflow identity, timing and high-level tool outcomes, never tool arguments, command strings, file contents or credentials.",
+ description: "Return a redacted live status snapshot for one workflow; arguments, file contents and credentials stay hidden.",
scope: "read",
annotations: { readOnlyHint: true, idempotentHint: true },
outputSchema: WORKFLOW_PROGRESS_OUTPUT,
@@ -23,7 +22,7 @@ export const WORKFLOW_LIFECYCLE_TOOLS: McpTool[] = [
},
limit: { key: "workflow.status", max: 30, windowMs: 60_000 },
inputSchema: S({
- workflow_id: { type: "string", description: "Exact id returned by workflow_start." },
+ workflow_id: { type: "string", description: "workflow_start id." },
}),
run: async (a, context) => {
const workflowId = str(a, "workflow_id");
@@ -36,16 +35,14 @@ export const WORKFLOW_LIFECYCLE_TOOLS: McpTool[] = [
},
{
name: "workflow_cancel",
- description:
- "Cancel one exact active workflow without saving a learned recipe. Use this only when the task is being abandoned or the prior run was interrupted; " +
- "workflow_id is required so a different conversation cannot be cancelled accidentally.",
+ description: "Cancel one exact active workflow without saving a learned recipe.",
scope: "write",
annotations: { idempotentHint: false },
limit: { key: "workflow.memory", max: 30, windowMs: 60_000 },
audit: { action: "workflow.cancel" as const, targetArg: "workflow_id" },
inputSchema: S({
- workflow_id: { type: "string", description: "Exact id returned by workflow_start." },
- reason: { type: "string", description: "Optional concise reason; no secrets or file contents." },
+ workflow_id: { type: "string", description: "workflow_start id." },
+ reason: { type: "string", description: "Optional reason; no secrets." },
}, ["workflow_id"]),
run: async (a, context) => {
const cancelled = await cancelWorkflow({
@@ -58,17 +55,15 @@ export const WORKFLOW_LIFECYCLE_TOOLS: McpTool[] = [
},
{
name: "workflow_finish",
- description:
- "Finish one exact learned workflow after verification. workflow_id is required so another conversation using the same token cannot close the wrong run. " +
- "MSO saves a redacted Evidence Receipt, task/failure memory, recipe quality, and safe automation-promotion metadata. New HIGH-risk workflows cannot claim success without explicit evidence.",
+ description: "Finish one verified workflow. MSO saves redacted evidence, memory, recipe quality and safe automation-learning metadata.",
scope: "write",
annotations: { idempotentHint: false },
limit: { key: "workflow.memory", max: 30, windowMs: 60_000 },
audit: { action: "workflow.finish" as const, targetArg: "workflow_id" },
inputSchema: S({
- workflow_id: { type: "string", description: "Exact id returned by workflow_start." },
- summary: { type: "string", description: "Concise outcome and verification result; no secrets." },
- success: { type: "boolean", description: "True only after the requested result is verified." },
+ workflow_id: { type: "string", description: "workflow_start id." },
+ summary: { type: "string", description: "Outcome and verification; no secrets." },
+ success: { type: "boolean", description: "True only after verification." },
evidence: evidenceSchema(),
}, ["workflow_id", "summary", "success"]),
run: async (a, context) => {
diff --git a/lib/mcp/tools-workflow-start.ts b/lib/mcp/tools-workflow-start.ts
index 8b2af407..5b8f9b20 100644
--- a/lib/mcp/tools-workflow-start.ts
+++ b/lib/mcp/tools-workflow-start.ts
@@ -17,10 +17,7 @@ import { optionalStringList, visibleTools, WORKFLOW_PROGRESS_OUTPUT, workflowPro
export const WORKFLOW_START_TOOL: McpTool =
{
name: "workflow_start",
- description:
- "The ONE startup call for a multi-step task. It starts the workflow, searches trusted skills and prior recipes, " +
- "resolves project aliases, reports the current toolset/version, and inspects repository context when available. " +
- "Do not call skills_search first for the same task; this already includes it. Multiple conversations may start isolated workflows on one token; correlate every later step with the returned workflow_id. This tool is orchestration-only and never opens a UI by itself.",
+ description: "Start one isolated multi-step workflow, resolve project context, and search trusted skills/recipes/graphs. Use the returned workflow_id on later steps.",
scope: "write",
annotations: { idempotentHint: false },
outputSchema: WORKFLOW_PROGRESS_OUTPUT,
@@ -31,11 +28,11 @@ export const WORKFLOW_START_TOOL: McpTool =
limit: { key: "workflow.memory", max: 30, windowMs: 60_000 },
audit: { action: "workflow.start" as const, targetArg: "project" },
inputSchema: S({
- intent: { type: "string", description: "The user's task in one complete sentence." },
- project: { type: "string", description: "Optional project id from projects_list, absolute path, name or alias (e.g. os-vps, mso, projects/mso)." },
- constraints: { type: "string", description: "Optional important constraints, such as no downtime or WebP only." },
- affected_paths: { type: "array", maxItems: 80, items: { type: "string" }, description: "Optional expected files/directories this workflow may touch; used for pre-merge collision detection." },
- reserved_resources: { type: "array", maxItems: 40, items: { type: "string" }, description: "Optional shared resources such as port:4173, database:development, deployment:staging, queue:name." },
+ intent: { type: "string", description: "Task in one sentence." },
+ project: { type: "string", description: "Optional project id/path/name/alias." },
+ constraints: { type: "string", description: "Optional constraints." },
+ affected_paths: { type: "array", maxItems: 80, items: { type: "string" }, description: "Optional touched paths." },
+ reserved_resources: { type: "array", maxItems: 40, items: { type: "string" }, description: "Optional shared resources." },
}, ["intent"]),
run: async (a, context) => {
const actor = context.workflowActor ?? context.actor;
diff --git a/lib/orchestration/capability-catalog.mjs b/lib/orchestration/capability-catalog.mjs
index c72621db..893b3bea 100644
--- a/lib/orchestration/capability-catalog.mjs
+++ b/lib/orchestration/capability-catalog.mjs
@@ -1,5 +1,5 @@
const ROUTES = [
- { id: "automation-flow", priority: 105, patterns: [/\b(automation flow|workflow graph|workflow|flow_catalog|flow_run|mso flow|reusable flow|alur otomatis|otomasi provider|n8n)\b/i], tools: ["workflow_graph_catalog", "workflow_graph_run", "workflow_graph_status", "workflow_graph_manage", "flow_catalog", "flow_run", "flow_status", "flow_manage", "integration_query"], budget: 12_000 },
+ { id: "automation-flow", priority: 105, patterns: [/\b(automation flow|workflow graph|workflow|flow_catalog|flow_run|mso flow|reusable flow|alur otomatis|otomasi provider|n8n)\b/i], tools: ["workflow_graph", "flow_catalog", "flow_run", "flow_status", "flow_manage", "integration_query"], budget: 12_000 },
{ id: "project-mcp-add", priority: 105, patterns: [/\b(add|connect|tambah|pasang)\b.{0,30}\bmcp\b/i], tools: ["project_mcp_manage", "project_mcp_tools", "integration_query", "integration_setup_open"], budget: 10_000 },
{ id: "project-asset", priority: 105, patterns: [/\b(attach|save|kirim|simpan)\b.{0,40}\b(asset|image|gambar)\b.{0,30}\b(project|proyek)\b/i], tools: ["session_artifacts", "project_asset_attach"], budget: 8_000 },
{ id: "managed-app-diagnostics", priority: 100, patterns: [/\b(hermes|openclaw|managed app|service)\b.*\b(down|error|crash|logs?|journal|status)\b/i, /\b(logs?|journal)\b.*\b(hermes|openclaw|service)\b/i], tools: ["apps_list", "apps_logs"], budget: 8_000 },