new wf_server: local static workflow server; docs

docs/current_roadmap.md

@@ -85,9 +85,11 @@ implementation state.
      [2026-06-03 long-lived workflow API boundary](./superpowers/specs/2026-06-03-long-lived-workflow-api-boundary.md):
      first slice should prove a lightweight local/static server that constructs
      `WorkflowApi` without `WfMcpService`; later slices add remote CLI targeting,
-     swappable HTTP/JSON-RPC/WebSocket/MCP transport adapters, source providers,
+     JSON-RPC-over-HTTP first transport, remote CLI targeting, WebSocket/MCP
-     auth, streaming/progress, transactional storage, and live upstream MCP
+     transport siblings, source providers, auth, streaming/progress,
-     sources.
+     transactional storage, and live upstream MCP sources.
+   - First slice implemented: `wf_server` can construct a local/static durable
+     `WorkflowApi` without `WfMcpService`. Transport adapters remain future work.
 
 5. **CLI/API alignment**
    - Let the CLI target either local process-backed stores/runtime or the future

docs/superpowers/plans/2026-06-03-local-static-workflow-server.md

@@ -1142,9 +1142,13 @@ Verification:
 These are intentionally not part of this implementation plan:
 
 1. **Transport adapter plan**
-   - Add `wf_transport_http` or equivalent.
+   - Add JSON-RPC 2.0 over HTTP, likely `wf_transport_rpc_http`.
-   - Start with health, run deployment, inspect run, read trace, resume run.
+   - Prefer `fastapi-jsonrpc` + `uvicorn` for server dispatch/docs.
-   - Keep routes thin over `WorkflowApi`.
+   - Start with health, list/inspect capabilities, run deployment, inspect run,
+     read trace, and resume run.
+   - Use stable dotted method names such as `workflow.runs.start`.
+   - Keep method handlers thin over `WorkflowApi`.
+   - Do not dynamically register saved workflows as JSON-RPC methods.
 
 2. **CLI remote target plan**
    - Let `wf_cli` choose local server composition or remote transport client.

docs/superpowers/specs/2026-06-03-long-lived-workflow-api-boundary.md

@@ -34,15 +34,11 @@ wf_server
   source/catalog/runtime/event implementation wiring
   durable WorkflowOperationContext construction
 
-wf_transport_http
+wf_transport_rpc_http
-  HTTP routes/controllers only
+  JSON-RPC 2.0 over HTTP endpoint/controller
-  request/response framework models
+  request/response envelope translation
   auth/session/streaming transport policy
 
-wf_transport_rpc
-  optional JSON-RPC over HTTP/WebSocket adapter
-  request/response envelope translation only
-
 wf_api
   protocol-neutral workflow application operations
   no HTTP imports
@@ -52,8 +48,8 @@ wf_mcp
   MCP transport and upstream MCP integration
 ```
 
-`wf_transport_http` should call `WorkflowApi` through the server composition. It
+`wf_transport_rpc_http` should call `WorkflowApi` through the server
-should not call `WfMcpService`.
+composition. It should not call `WfMcpService`.
 
 HTTP is one transport adapter, not "the API." JSON-RPC over HTTP, JSON-RPC over
 WebSocket, and possibly MCP can become sibling transports around the same
@@ -94,6 +90,15 @@ It should prove:
 - deployment run/inspect/trace/resume semantics match local behavior
 - no direct dependency on `WfMcpService`
 
+Implementation status:
+
+- `wf_server.build_local_static_workflow_server()` constructs a durable
+  `WorkflowApi` with required file-backed stores, local `wf.std`/`wf.recipes`
+  sources, and a local runtime runner.
+- This first slice has no transport adapter. Clients still call the in-process
+  `WorkflowApi` in tests; HTTP/JSON-RPC/WebSocket/MCP transport adapters are
+  later slices.
+
 First slice should not include:
 
 - live upstream MCP source management
@@ -231,11 +236,35 @@ Do not move large service sets in the first slice.
 
 ## Later Slice Pointers
 
-### Slice 2: HTTP Transport Adapter
+### Slice 2: JSON-RPC HTTP Transport Adapter
 
-Add the first transport package, likely `wf_transport_http`.
+Add the first transport package as JSON-RPC 2.0 over HTTP, likely
+`wf_transport_rpc_http`.
 
-This slice should expose a small route set over the existing server composition:
+This is preferred over REST for the first remote CLI/server path because CLI and
+agent clients want stable operation names more than resource-shaped URLs. The
+method names should be dotted strings, matching the mental model already used by
+MCP/admin tool names without inheriting MCP's dynamic tool-list behavior.
+
+Proposed initial method names:
+
+```text
+workflow.capabilities.list
+workflow.capabilities.inspect
+workflow.drafts.create_from_capability
+workflow.drafts.patch
+workflow.drafts.validate
+workflow.artifacts.save
+workflow.deployments.save
+workflow.deployments.validate
+workflow.runs.start
+workflow.runs.inspect
+workflow.runs.trace
+workflow.runs.resume
+```
+
+This slice should expose a small method set over the existing server
+composition:
 
 - health/status
 - list/inspect capabilities
@@ -246,6 +275,32 @@ This slice should expose a small route set over the existing server composition:
 
 It should not implement source provider management yet.
 
+Preferred implementation dependency:
+
+```bash
+uv add fastapi-jsonrpc uvicorn
+```
+
+`fastapi-jsonrpc` is the recommended server library for this slice because it
+keeps JSON-RPC 2.0 dispatch, errors, and docs near FastAPI/Pydantic instead of
+requiring a local hand-rolled dispatcher.
+
+Client-side CLI code can use `httpx` directly for now. A typed client wrapper
+can be added when the method set stabilizes.
+
+Guardrails:
+
+- Do not dynamically register saved workflows as JSON-RPC methods.
+- Do not add a JSON-RPC equivalent of MCP `tools/list` as the primary execution
+  surface.
+- Do not require client session state to run or resume workflows.
+- Keep durable state addressed by explicit ids: `artifact_id`, `deployment_id`,
+  `run_id`.
+- Keep trace reads bounded by explicit range.
+- Define request/response models explicitly with Pydantic; do not pass raw
+  arbitrary dicts through the transport layer when a stable request shape is
+  known.
+
 ### Slice 3: CLI Remote Target
 
 Allow `wf_cli` to target either:
@@ -255,14 +310,13 @@ Allow `wf_cli` to target either:
 
 The CLI command surface should stay stable. Only context construction changes.
 
-### Slice 4: JSON-RPC / WebSocket Transport
+### Slice 4: WebSocket Transport
 
-If HTTP REST starts becoming awkward for agents or streaming, add JSON-RPC as a
+If JSON-RPC over HTTP starts becoming awkward for streaming/progress, add a
-transport sibling rather than changing `WorkflowApi`.
+WebSocket transport sibling rather than changing `WorkflowApi`.
 
 Possible shapes:
 
-- JSON-RPC over HTTP
 - JSON-RPC over WebSocket
 - future MCP transport over the same server-owned `WorkflowApi`
 
@@ -326,10 +380,9 @@ This needs explicit policies for:
 
 1. Package name: `wf_server` is the recommended process-composition package,
    but the exact name can change before implementation.
-2. HTTP framework: likely FastAPI for `wf_transport_http`, but the first server
+2. HTTP framework: likely FastAPI under `fastapi-jsonrpc`.
-   slice does not require choosing until route implementation starts.
+3. RPC framework: use `fastapi-jsonrpc` for the first JSON-RPC-over-HTTP slice
-3. RPC framework: undecided. JSON-RPC should be added only if it is clearly
+   unless evaluation finds a blocking issue.
-   better for agent clients or streaming/progress.
 4. Source-provider extraction: some reusable code currently lives in
    `wf_mcp.broker.service`; first slice should avoid moving it unless a small
    dependency-free helper is obviously needed.

src/wf_api/__init__.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 from .listing import matches_query, paged_list_payload
 from .artifacts import WorkflowArtifactApi
+from .local_sources import builtin_sources, get_qualified_spec, qualify_spec
 from .models import RawWorkflowPlan, TraceRange
 from .capabilities import WorkflowCapabilityApi
 from .constants import (
@@ -43,8 +44,11 @@ from .durable_context import durable_workflow_api, require_workflow_stores
 
 __all__ = [
     "DEFAULT_CALL_STEP_ID",
+    "builtin_sources",
+    "get_qualified_spec",
     "matches_query",
     "paged_list_payload",
+    "qualify_spec",
     "DEFAULT_ERROR_OUTCOME",
     "DEFAULT_ERROR_STEP_ID",
     "DEFAULT_OK_OUTCOME",

src/wf_api/local_sources.py

@@ -0,0 +1,186 @@
+from __future__ import annotations
+
+from collections.abc import Mapping
+from typing import TYPE_CHECKING, Any
+
+from wf_authoring import NodeSpec, coalesce, concat, constant, default_if_none
+from wf_authoring import extract_field, filter_items, filter_items_present, first_item
+from wf_authoring import first_item_maybe, first_item_or_none, is_empty, last_item
+from wf_authoring import last_item_or_none, length, node, pick_key, pick_path
+from wf_authoring import project_fields, rename_fields, runtime_error, truthy
+from wf_authoring import extract_text_content
+from wf_core.runtime.ops.merges import DEFAULT_REDUCER_DEFINITIONS
+from wf_platform import (
+    CapabilityBuckets,
+    CapabilitySource,
+    SourcePermissions,
+    SourceVisibility,
+)
+
+if TYPE_CHECKING:
+    from wf_core import ReducerSpec
+
+BUILTIN_SOURCE_ID = "wf.std"
+"""Internal source id for workflow standard-library node specs."""
+
+BUILTIN_CONNECTION_ID = BUILTIN_SOURCE_ID
+"""Compatibility alias for older MCP broker code."""
+
+MCP_SOURCE_ID = "wf.mcp"
+"""Reserved source id for future workflow-safe MCP utility node specs."""
+
+RECIPE_SOURCE_ID = "wf.recipes"
+"""Internal source id for first-party composed workflow recipes."""
+
+AUTHORING_STD_SPECS: tuple[NodeSpec[Any, Any], ...] = (
+    coalesce,
+    default_if_none,
+    constant,
+    pick_key,
+    pick_path,
+    project_fields,
+    rename_fields,
+    truthy,
+    runtime_error,
+    first_item,
+    first_item_or_none,
+    first_item_maybe,
+    last_item,
+    last_item_or_none,
+    length,
+    is_empty,
+    filter_items,
+    filter_items_present,
+    extract_field,
+    concat,
+)
+"""Existing authoring ops exposed through the workflow stdlib."""
+
+RECIPE_SPECS: tuple[NodeSpec[Any, Any], ...] = (extract_text_content,)
+"""Composed first-party recipes exposed as workflow-facing capabilities."""
+
+
+def qualify_node_name(source_id: str, local_name: str) -> str:
+    """Return one source-qualified node name without assuming MCP connections."""
+    if not source_id:
+        raise ValueError("source_id must not be empty")
+    if not local_name:
+        raise ValueError("local node name must not be empty")
+    return f"{source_id}.{local_name}"
+
+
+def qualify_spec(source_id: str, spec: NodeSpec[Any, Any]) -> NodeSpec[Any, Any]:
+    """Return a copy of a spec with its node name scoped to a source."""
+    return NodeSpec(
+        name=qualify_node_name(source_id, spec.name),
+        input_model=spec.input_model,
+        output_model=spec.output_model,
+        outcomes=spec.outcomes,
+        fn=spec.fn,
+        description=spec.description,
+        is_async=spec.is_async,
+        accepts_context=spec.accepts_context,
+        input_schema_contract=spec.input_schema_contract,
+        output_schema_contract=spec.output_schema_contract,
+    )
+
+
+def get_qualified_spec(
+    sources: Mapping[str, CapabilitySource],
+    qualified_name: str,
+) -> NodeSpec[Any, Any]:
+    """Resolve a namespaced node spec from enabled planner-visible sources."""
+    for source in sources.values():
+        if not source.enabled or not source.visibility.planner:
+            continue
+        spec = source.capabilities.node_specs.get(qualified_name)
+        if spec is not None:
+            return spec
+    raise KeyError(f"unknown qualified node {qualified_name!r}")
+
+
+def _qualified_specs(
+    source_id: str,
+    specs: tuple[NodeSpec[Any, Any], ...],
+) -> dict[str, NodeSpec[Any, Any]]:
+    """Return specs with authoring names rewritten under one source id."""
+    local_specs = [
+        node(spec, name=spec.name.removeprefix("authoring.")) for spec in specs
+    ]
+    qualified_specs = [qualify_spec(source_id, spec) for spec in local_specs]
+    return {spec.name: spec for spec in qualified_specs}
+
+
+def builtin_specs() -> dict[str, NodeSpec[Any, Any]]:
+    """Return primitive built-in NodeSpecs available to raw workflow plans."""
+    return _qualified_specs(BUILTIN_SOURCE_ID, AUTHORING_STD_SPECS)
+
+
+def recipe_specs() -> dict[str, NodeSpec[Any, Any]]:
+    """Return composed first-party recipe specs."""
+    return _qualified_specs(RECIPE_SOURCE_ID, RECIPE_SPECS)
+
+
+def builtin_reducers() -> dict[str, ReducerSpec]:
+    """Return built-in reducers owned by the workflow standard library."""
+    return {
+        definition.spec.name: definition.spec
+        for definition in DEFAULT_REDUCER_DEFINITIONS.values()
+    }
+
+
+def builtin_reducer_definitions():
+    """Return executable built-in reducers for trusted runtime dependency wiring."""
+    return dict(DEFAULT_REDUCER_DEFINITIONS)
+
+
+def builtin_sources() -> dict[str, CapabilitySource]:
+    """Return all local workflow-facing capability sources."""
+    return {
+        BUILTIN_SOURCE_ID: CapabilitySource(
+            id=BUILTIN_SOURCE_ID,
+            kind="system",
+            capabilities=CapabilityBuckets(
+                node_specs=builtin_specs(),
+                reducers=builtin_reducers(),
+                reducer_definitions=builtin_reducer_definitions(),
+            ),
+            visibility=SourceVisibility(
+                planner=True,
+                mcp_client=True,
+                admin_dashboard=True,
+            ),
+            permissions=SourcePermissions(safe_for_workflow=True),
+            description="Workflow standard-library nodes.",
+        ),
+        RECIPE_SOURCE_ID: CapabilitySource(
+            id=RECIPE_SOURCE_ID,
+            kind="system",
+            capabilities=CapabilityBuckets(node_specs=recipe_specs()),
+            visibility=SourceVisibility(
+                planner=True,
+                mcp_client=True,
+                admin_dashboard=True,
+            ),
+            permissions=SourcePermissions(safe_for_workflow=True),
+            description="First-party workflow recipes composed from standard nodes.",
+        ),
+    }
+
+
+__all__ = [
+    "AUTHORING_STD_SPECS",
+    "BUILTIN_CONNECTION_ID",
+    "BUILTIN_SOURCE_ID",
+    "MCP_SOURCE_ID",
+    "RECIPE_SOURCE_ID",
+    "RECIPE_SPECS",
+    "builtin_reducer_definitions",
+    "builtin_reducers",
+    "builtin_sources",
+    "builtin_specs",
+    "get_qualified_spec",
+    "qualify_node_name",
+    "qualify_spec",
+    "recipe_specs",
+]

src/wf_mcp/broker/service/builtins.py

@@ -1,133 +1,41 @@
+"""Compatibility exports for workflow local sources.
+
+Canonical local workflow source helpers live in `wf_api.local_sources` so
+non-MCP process hosts can construct `wf.std` without importing broker internals.
+"""
+
 from __future__ import annotations
 
-from typing import TYPE_CHECKING, Any
+from wf_api.local_sources import (
-
+    AUTHORING_STD_SPECS,
-from wf_authoring import NodeSpec, coalesce, concat, constant, default_if_none
+    BUILTIN_CONNECTION_ID,
-from wf_authoring import extract_field, filter_items, filter_items_present, first_item
+    BUILTIN_SOURCE_ID,
-from wf_authoring import first_item_maybe, first_item_or_none, is_empty, last_item
+    MCP_SOURCE_ID,
-from wf_authoring import last_item_or_none, length, node, pick_key, pick_path
+    RECIPE_SOURCE_ID,
-from wf_authoring import project_fields, rename_fields, runtime_error, truthy
+    RECIPE_SPECS,
-from wf_authoring import extract_text_content
+    builtin_reducer_definitions,
-from wf_core.runtime.ops.merges import DEFAULT_REDUCER_DEFINITIONS
+    builtin_reducers,
-
+    builtin_sources,
-from wf_platform import (
+    builtin_specs,
-    CapabilityBuckets,
+    get_qualified_spec,
-    CapabilitySource,
+    qualify_node_name,
-    SourcePermissions,
+    qualify_spec,
-    SourceVisibility,
+    recipe_specs,
 )
-from wf_mcp.broker.service.specs import qualify_spec
 
-if TYPE_CHECKING:
+__all__ = [
-    from wf_core import ReducerSpec
+    "AUTHORING_STD_SPECS",
-
+    "BUILTIN_CONNECTION_ID",
-BUILTIN_CONNECTION_ID = "wf.std"
+    "BUILTIN_SOURCE_ID",
-"""Internal source id for workflow standard-library node specs."""
+    "MCP_SOURCE_ID",
-
+    "RECIPE_SOURCE_ID",
-MCP_SOURCE_ID = "wf.mcp"
+    "RECIPE_SPECS",
-"""Reserved source id for future workflow-safe MCP utility node specs."""
+    "builtin_reducer_definitions",
-
+    "builtin_reducers",
-RECIPE_SOURCE_ID = "wf.recipes"
+    "builtin_sources",
-"""Internal source id for first-party composed workflow recipes."""
+    "builtin_specs",
-
+    "get_qualified_spec",
-
+    "qualify_node_name",
-AUTHORING_STD_SPECS: tuple[NodeSpec[Any, Any], ...] = (
+    "qualify_spec",
-    coalesce,
+    "recipe_specs",
-    default_if_none,
+]
-    constant,
-    pick_key,
-    pick_path,
-    project_fields,
-    rename_fields,
-    truthy,
-    runtime_error,
-    first_item,
-    first_item_or_none,
-    first_item_maybe,
-    last_item,
-    last_item_or_none,
-    length,
-    is_empty,
-    filter_items,
-    filter_items_present,
-    extract_field,
-    concat,
-)
-"""Existing authoring ops that are also exposed through the workflow stdlib."""
-
-
-RECIPE_SPECS: tuple[NodeSpec[Any, Any], ...] = (extract_text_content,)
-"""Composed first-party recipes exposed as capabilities."""
-
-
-def _qualified_specs(
-    source_id: str,
-    specs: tuple[NodeSpec[Any, Any], ...],
-) -> dict[str, NodeSpec[Any, Any]]:
-    """Return specs with authoring names rewritten under one source id."""
-    local_specs = [
-        node(spec, name=spec.name.removeprefix("authoring.")) for spec in specs
-    ]
-    qualified_specs = [qualify_spec(source_id, spec) for spec in local_specs]
-    return {spec.name: spec for spec in qualified_specs}
-
-
-def builtin_specs() -> dict[str, NodeSpec[Any, Any]]:
-    """Return primitive built-in NodeSpecs available to raw broker workflow plans."""
-    return _qualified_specs(BUILTIN_CONNECTION_ID, AUTHORING_STD_SPECS)
-
-
-def recipe_specs() -> dict[str, NodeSpec[Any, Any]]:
-    """Return composed first-party recipe specs.
-
-    Recipes are wrapper-node subgraphs today. They are useful workflow-facing
-    capabilities, but parent runs do not yet see their child graph frames.
-    """
-    return _qualified_specs(RECIPE_SOURCE_ID, RECIPE_SPECS)
-
-
-def builtin_reducers() -> dict[str, ReducerSpec]:
-    """Return built-in reducers owned by the workflow standard library."""
-    return {
-        definition.spec.name: definition.spec
-        for definition in DEFAULT_REDUCER_DEFINITIONS.values()
-    }
-
-
-def builtin_reducer_definitions():
-    """Return executable built-in reducers for trusted runtime dependency wiring."""
-    return dict(DEFAULT_REDUCER_DEFINITIONS)
-
-
-def builtin_sources() -> dict[str, CapabilitySource]:
-    """Return all broker-local capability sources."""
-    return {
-        BUILTIN_CONNECTION_ID: CapabilitySource(
-            id=BUILTIN_CONNECTION_ID,
-            kind="system",
-            capabilities=CapabilityBuckets(
-                node_specs=builtin_specs(),
-                reducers=builtin_reducers(),
-                reducer_definitions=builtin_reducer_definitions(),
-            ),
-            visibility=SourceVisibility(
-                planner=True,
-                mcp_client=True,
-                admin_dashboard=True,
-            ),
-            permissions=SourcePermissions(safe_for_workflow=True),
-            description="Workflow standard-library nodes.",
-        ),
-        RECIPE_SOURCE_ID: CapabilitySource(
-            id=RECIPE_SOURCE_ID,
-            kind="system",
-            capabilities=CapabilityBuckets(node_specs=recipe_specs()),
-            visibility=SourceVisibility(
-                planner=True,
-                mcp_client=True,
-                admin_dashboard=True,
-            ),
-            permissions=SourcePermissions(safe_for_workflow=True),
-            description="First-party workflow recipes composed from standard nodes.",
-        ),
-    }

src/wf_server/__init__.py

@@ -0,0 +1,19 @@
+from __future__ import annotations
+
+from .context import (
+    InMemoryWorkflowEventRecorder,
+    LocalWorkflowRuntimeRunner,
+    StaticWorkflowSpecProvider,
+    WorkflowServer,
+    WorkflowServerConfig,
+    build_local_static_workflow_server,
+)
+
+__all__ = [
+    "InMemoryWorkflowEventRecorder",
+    "LocalWorkflowRuntimeRunner",
+    "StaticWorkflowSpecProvider",
+    "WorkflowServer",
+    "WorkflowServerConfig",
+    "build_local_static_workflow_server",
+]

src/wf_server/context.py

@@ -0,0 +1,273 @@
+from __future__ import annotations
+
+from collections.abc import Mapping
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Any
+
+from wf_api import WorkflowApi, durable_workflow_api
+from wf_api.local_sources import builtin_sources, get_qualified_spec
+from wf_api.models import RawWorkflowPlan, TraceRange
+from wf_api.operation_context import (
+    WorkflowEventRecorder,
+    WorkflowOperationContext,
+    WorkflowRuntimeRunner,
+    WorkflowSpecProvider,
+)
+from wf_api.runtime_dependencies import resolve_runtime_dependencies
+from wf_api.saved_subgraphs import (
+    SavedSubgraphTree,
+    prepare_saved_subgraphs,
+    resolve_saved_subgraph_tree,
+)
+from wf_api.stores import WorkflowStores, file_workflow_stores
+from wf_artifacts import WorkflowArtifact, WorkflowDeployment
+from wf_authoring import NodeSpec
+from wf_core import (
+    NodeUse,
+    RunState,
+    Workflow,
+    execute_workflow_result_async,
+    resume_workflow_result_async,
+)
+from wf_platform import CapabilitySource
+
+
+@dataclass(frozen=True, slots=True)
+class WorkflowServerConfig:
+    """Configuration for the first local/static workflow server slice."""
+
+    store_root: Path
+
+
+@dataclass(slots=True)
+class InMemoryWorkflowEventRecorder(WorkflowEventRecorder):
+    """Small process-local event sink for server composition tests."""
+
+    events: list[dict[str, Any]] = field(default_factory=list)
+
+    def record_event(self, event: object) -> None:
+        self.events.append({"kind": "adapter_event", "event": event})
+
+    def record_workflow_event(
+        self,
+        event_type: str,
+        *,
+        capability_id: str,
+        payload: dict[str, Any],
+    ) -> None:
+        self.events.append(
+            {
+                "kind": event_type,
+                "capability_id": capability_id,
+                "payload": payload,
+            }
+        )
+
+
+@dataclass(frozen=True, slots=True)
+class StaticWorkflowSpecProvider(WorkflowSpecProvider):
+    """Source provider for local/static server capabilities."""
+
+    sources: Mapping[str, CapabilitySource]
+
+    @property
+    def capability_sources(self) -> dict[str, CapabilitySource]:
+        return dict(self.sources)
+
+    def get_qualified_spec(self, qualified_name: str) -> NodeSpec[Any, Any]:
+        return get_qualified_spec(self.sources, qualified_name)
+
+
+@dataclass(slots=True)
+class LocalWorkflowRuntimeRunner(WorkflowRuntimeRunner):
+    """Run workflow plans against local/static source catalogs."""
+
+    specs: StaticWorkflowSpecProvider
+    artifact_store: Any
+
+    def compile_plan(
+        self,
+        plan: RawWorkflowPlan,
+        node_name_bindings: dict[str, str] | None = None,
+    ) -> Workflow:
+        node_defs: dict[str, Any] = {}
+        bindings = node_name_bindings or {}
+        for step in plan.nodes:
+            if not isinstance(step, NodeUse):
+                continue
+            qualified_name = bindings.get(step.node, step.node)
+            spec = self.specs.get_qualified_spec(qualified_name)
+            node_defs[qualified_name] = spec.to_node_def()
+
+        nodes = []
+        for node in plan.nodes:
+            node_payload = node.model_dump(by_alias=True)
+            if isinstance(node, NodeUse):
+                node_payload["node"] = bindings.get(node.node, node.node)
+            nodes.append(node_payload)
+
+        return Workflow.model_validate(
+            {
+                "name": plan.name,
+                "input_schema": plan.input_schema,
+                "state_schema": plan.state_schema,
+                "output_schema": plan.output_schema,
+                "output": [binding.model_dump(mode="json") for binding in plan.output],
+                "outcomes": plan.outcomes,
+                "start": plan.start,
+                "node_defs": [node.model_dump() for node in node_defs.values()],
+                "nodes": nodes,
+                "edges": [edge.model_dump(by_alias=True) for edge in plan.edges],
+            }
+        )
+
+    def prepare_workflow_runtime(
+        self,
+        plan: RawWorkflowPlan,
+        *,
+        deployment: WorkflowDeployment | None,
+        artifact: WorkflowArtifact | None,
+        saved_subgraph_tree: SavedSubgraphTree | None = None,
+    ) -> tuple[Workflow, dict[str, Any], dict[str, Any], dict[str, Any]]:
+        plan_node_names = [
+            node.node for node in plan.nodes if isinstance(node, NodeUse)
+        ]
+        runtime_artifact = artifact or WorkflowArtifact(
+            id=plan.name,
+            version=1,
+            title=plan.name,
+            input_schema=plan.input_schema,
+            output_schema=plan.output_schema,
+            outcomes=("completed",),
+            plan=plan.model_dump(mode="json", by_alias=True),
+        )
+        dependencies = resolve_runtime_dependencies(
+            artifact=runtime_artifact,
+            deployment=deployment,
+            sources=self.specs.capability_sources,
+            plan_node_names=plan_node_names,
+        )
+        prepared_subgraphs = {}
+        if saved_subgraph_tree is not None:
+            prepared_subgraphs = prepare_saved_subgraphs(
+                tree=saved_subgraph_tree,
+                deployment=deployment,
+                sources=self.specs.capability_sources,
+                compile_plan=self.compile_plan,
+            )
+        elif artifact is not None and self.artifact_store is not None:
+            tree = resolve_saved_subgraph_tree(
+                root_artifact=artifact,
+                artifact_store=self.artifact_store,
+            )
+            prepared_subgraphs = prepare_saved_subgraphs(
+                tree=tree,
+                deployment=deployment,
+                sources=self.specs.capability_sources,
+                compile_plan=self.compile_plan,
+            )
+        workflow = self.compile_plan(plan, dependencies.node_name_bindings)
+        return (
+            workflow,
+            dependencies.node_registry,
+            dependencies.reducers,
+            prepared_subgraphs,
+        )
+
+    async def run_workflow_from_plan(
+        self,
+        plan: RawWorkflowPlan,
+        workflow_input: dict[str, Any],
+        deployment: WorkflowDeployment | None = None,
+        artifact: WorkflowArtifact | None = None,
+        saved_subgraph_tree: SavedSubgraphTree | None = None,
+    ) -> RunState:
+        workflow, registry, reducers, prepared_subgraphs = (
+            self.prepare_workflow_runtime(
+                plan,
+                deployment=deployment,
+                artifact=artifact,
+                saved_subgraph_tree=saved_subgraph_tree,
+            )
+        )
+        return await execute_workflow_result_async(
+            workflow,
+            workflow_input,
+            registry,
+            reducers=reducers,
+            subgraphs=prepared_subgraphs,
+        )
+
+    async def resume_workflow_from_plan(
+        self,
+        plan: RawWorkflowPlan,
+        run: RunState,
+        *,
+        resume_payload: dict[str, Any],
+        resume_outcome: str,
+        deployment: WorkflowDeployment | None = None,
+        artifact: WorkflowArtifact | None = None,
+        saved_subgraph_tree: SavedSubgraphTree | None = None,
+    ) -> RunState:
+        workflow, registry, reducers, prepared_subgraphs = (
+            self.prepare_workflow_runtime(
+                plan,
+                deployment=deployment,
+                artifact=artifact,
+                saved_subgraph_tree=saved_subgraph_tree,
+            )
+        )
+        return await resume_workflow_result_async(
+            workflow,
+            run,
+            registry,
+            resume_payload=resume_payload,
+            resume_outcome=resume_outcome,
+            reducers=reducers,
+            subgraphs=prepared_subgraphs,
+        )
+
+
+@dataclass(frozen=True, slots=True)
+class WorkflowServer:
+    """First-slice long-lived server composition without transport concerns."""
+
+    config: WorkflowServerConfig
+    stores: WorkflowStores
+    context: WorkflowOperationContext
+    api: WorkflowApi
+    events: InMemoryWorkflowEventRecorder
+
+    @staticmethod
+    def trace_range(*, start: int, limit: int) -> TraceRange:
+        return TraceRange(start=start, limit=limit)
+
+
+def build_local_static_workflow_server(root: str | Path) -> WorkflowServer:
+    """Build a durable local/static workflow server composition."""
+    config = WorkflowServerConfig(store_root=Path(root))
+    stores = file_workflow_stores(config.store_root)
+    events = InMemoryWorkflowEventRecorder()
+    specs = StaticWorkflowSpecProvider(builtin_sources())
+    runtime = LocalWorkflowRuntimeRunner(
+        specs=specs,
+        artifact_store=stores.artifact_store,
+    )
+    context = WorkflowOperationContext(
+        artifact_store=stores.artifact_store,
+        draft_workspace_store=stores.draft_workspace_store,
+        run_store=stores.run_store,
+        events=events,
+        specs=specs,
+        runtime=runtime,
+        live_sources=None,
+    )
+    api = durable_workflow_api(context)
+    return WorkflowServer(
+        config=config,
+        stores=stores,
+        context=context,
+        api=api,
+        events=events,
+    )

tests/wf_api/test_local_sources.py

@@ -0,0 +1,45 @@
+from __future__ import annotations
+
+from wf_api.local_sources import (
+    BUILTIN_SOURCE_ID,
+    RECIPE_SOURCE_ID,
+    builtin_sources,
+    get_qualified_spec,
+    qualify_spec,
+)
+from wf_authoring import constant
+
+
+def test_builtin_sources_expose_workflow_stdlib() -> None:
+    sources = builtin_sources()
+
+    assert BUILTIN_SOURCE_ID == "wf.std"
+    assert RECIPE_SOURCE_ID == "wf.recipes"
+    assert "wf.std" in sources
+    assert "wf.std.constant" in sources["wf.std"].capabilities.node_specs
+    assert "wf.std.replace" in sources["wf.std"].capabilities.reducers
+
+
+def test_get_qualified_spec_resolves_planner_visible_spec() -> None:
+    sources = builtin_sources()
+
+    spec = get_qualified_spec(sources, "wf.std.constant")
+
+    assert spec.name == "wf.std.constant"
+    assert spec.outcomes == ("ok",)
+
+
+def test_qualify_spec_scopes_authoring_node_name() -> None:
+    qualified = qualify_spec("custom.local", constant)
+
+    assert qualified.name == "custom.local.authoring.constant"
+    assert qualified.input_model is constant.input_model
+    assert qualified.output_model is constant.output_model
+
+
+def test_mcp_builtin_module_reexports_canonical_helpers() -> None:
+    from wf_mcp.broker.service import builtins as mcp_builtins
+
+    assert mcp_builtins.BUILTIN_CONNECTION_ID == BUILTIN_SOURCE_ID
+    assert mcp_builtins.BUILTIN_SOURCE_ID == BUILTIN_SOURCE_ID
+    assert mcp_builtins.builtin_sources is builtin_sources

tests/wf_server/test_local_static_server.py

@@ -0,0 +1,159 @@
+from __future__ import annotations
+
+import ast
+import asyncio
+from pathlib import Path
+
+from wf_api.models import RawWorkflowPlan
+from wf_core import END
+from wf_server import build_local_static_workflow_server
+
+
+def _constant_plan() -> RawWorkflowPlan:
+    return RawWorkflowPlan.model_validate(
+        {
+            "name": "server_constant",
+            "input_schema": {"type": "object", "properties": {}},
+            "state_schema": {
+                "type": "object",
+                "properties": {
+                    "result": {"type": "string", "reducer": "wf.std.replace"}
+                },
+            },
+            "output_schema": {
+                "type": "object",
+                "properties": {"result": {"type": "string"}},
+                "required": ["result"],
+            },
+            "outcomes": ["ok"],
+            "start": "constant",
+            "nodes": [
+                {
+                    "id": "constant",
+                    "type": "node",
+                    "node": "wf.std.constant",
+                    "input": [
+                        {
+                            "value": "hello from server",
+                            "target": {"root": "local", "parts": ["value"]},
+                        }
+                    ],
+                    "output": [
+                        {
+                            "source": {"root": "local", "parts": ["value"]},
+                            "target": {"root": "state", "parts": ["result"]},
+                        }
+                    ],
+                }
+            ],
+            "edges": [{"from": "constant", "outcome": "ok", "to": END}],
+            "output": [
+                {
+                    "path": {"root": "state", "parts": ["result"]},
+                    "target": {"root": "local", "parts": ["result"]},
+                }
+            ],
+        }
+    )
+
+
+def test_wf_server_context_imports_no_wfmcp_service() -> None:
+    path = Path("src/wf_server/context.py")
+    tree = ast.parse(path.read_text(encoding="utf-8"), filename=str(path))
+
+    violations: list[str] = []
+    for node in ast.walk(tree):
+        if isinstance(node, ast.ImportFrom) and node.module is not None:
+            if node.module == "wf_mcp.broker" or node.module.endswith(".core"):
+                violations.append(f"{node.lineno}: from {node.module} import ...")
+        elif isinstance(node, ast.Import):
+            for alias in node.names:
+                if alias.name in {"wf_mcp.broker", "wf_mcp.broker.service.core"}:
+                    violations.append(f"{node.lineno}: import {alias.name}")
+
+    assert violations == []
+
+
+def test_local_static_server_runs_deployment_and_persists_run(tmp_path) -> None:
+    server = build_local_static_workflow_server(tmp_path / "store")
+    api = server.api
+    plan = _constant_plan()
+
+    artifact_result = asyncio.run(
+        api.create_artifact_from_plan(
+            artifact_id="server_constant",
+            version=1,
+            title="Server Constant",
+            plan=plan,
+            outcomes=["ok"],
+            source_bindings={"wf.std": "wf.std"},
+        )
+    )
+    deployment_result = asyncio.run(
+        api.save_deployment(
+            {
+                "id": "server_constant.default",
+                "artifact_id": "server_constant",
+                "artifact_version": 1,
+                "bindings": [{"logical_source": "wf.std", "concrete_source": "wf.std"}],
+            }
+        )
+    )
+    run_result = asyncio.run(
+        api.run_deployment(
+            deployment_id="server_constant.default",
+            workflow_input={},
+        )
+    )
+
+    assert artifact_result["artifact_id"] == "server_constant"
+    assert deployment_result["deployment_id"] == "server_constant.default"
+    assert run_result["status"] == "completed"
+    assert run_result["output"]["result"] == "hello from server"
+    assert isinstance(run_result["run_id"], str)
+    assert (
+        server.stores.run_store.get_run(run_result["run_id"]).id == run_result["run_id"]
+    )
+
+
+def test_local_static_server_inspects_and_reads_bounded_trace(tmp_path) -> None:
+    server = build_local_static_workflow_server(tmp_path / "store")
+    api = server.api
+    plan = _constant_plan()
+    asyncio.run(
+        api.create_artifact_from_plan(
+            artifact_id="server_trace",
+            version=1,
+            title="Server Trace",
+            plan=plan.model_copy(update={"name": "server_trace"}),
+            outcomes=["ok"],
+            source_bindings={"wf.std": "wf.std"},
+        )
+    )
+    asyncio.run(
+        api.save_deployment(
+            {
+                "id": "server_trace.default",
+                "artifact_id": "server_trace",
+                "artifact_version": 1,
+                "bindings": [{"logical_source": "wf.std", "concrete_source": "wf.std"}],
+            }
+        )
+    )
+    run_result = asyncio.run(
+        api.run_deployment(deployment_id="server_trace.default", workflow_input={})
+    )
+
+    summary = asyncio.run(api.inspect_run(run_id=run_result["run_id"]))
+    trace = asyncio.run(
+        api.read_run_trace(
+            run_id=run_result["run_id"],
+            trace_range=server.trace_range(start=0, limit=1),
+        )
+    )
+
+    assert "trace" not in summary
+    assert summary["trace_count"] >= 1
+    assert trace["trace_start"] == 0
+    assert trace["trace_limit"] == 1
+    assert len(trace["trace"]) == 1