cleveragents/cleveragents-core

Fork 3

UAT: MCP adapter infers resource slots but never wires them into domain Tool objects or DB #5474

New issue

Open

opened 2026-04-09 06:57:08 +00:00 by HAL9000 · 1 comment

HAL9000 commented

2026-04-09 06:57:08 +00:00

Owner

Bug Report

Feature Area: MCP Adapter — Tool call routing / Resource slot binding
Severity: Backlog — resource slot inference is implemented but not consumed
Found by: UAT Testing (tool-router-mcp-adapter area)

What Was Tested

Code-level analysis of src/cleveragents/mcp/adapter.py — specifically the register_tools() method (lines 549–638) and the infer_resource_slots() static method (lines 652–736).

Expected Behavior (from spec)

When MCP tools are registered into the ToolRegistry, their inferred resource slots should be stored in the domain Tool objects and persisted to the database via the tool_resource_bindings table (migration c1_001). This enables the runtime to correctly bind resources to MCP tool invocations.

Actual Behavior

The register_tools() method infers resource slots and stores them in source_metadata, but there is a TODO comment acknowledging that nothing downstream reads them:

# TODO(#882): resource_slots are stored in source_metadata but
# nothing downstream reads them yet.  The ToolRegistry persists
# resource bindings via domain Tool.resource_slots, and the DB
# migration (c1_001) stores them in tool_resource_bindings.
# A follow-up ticket should wire inferred slots into the domain
# Tool objects so the registry and DB actually consume them.
cap_meta = self.capability_metadata
spec = ToolSpec(
    ...
    source_metadata={
        "server": self._config.name,
        "resource_slots": slot_dicts,  # Stored but never read
        ...
    },
)

The ToolRegistry.find_tools_for_resource() method looks for source_metadata["resource_bindings"] (not "resource_slots"), so even the key name is inconsistent.

Code Location

Bug: src/cleveragents/mcp/adapter.py, lines 611–616 (TODO comment)
Inconsistency: src/cleveragents/tool/registry.py, line 127 — looks for "resource_bindings" but MCP stores "resource_slots"

Impact

ToolRegistry.find_tools_for_resource() will never find MCP tools even when they have matching resource slots
MCP tool resource bindings are never persisted to the tool_resource_bindings DB table
The resource slot inference logic in infer_resource_slots() is effectively dead code

Fix

Rename the key from "resource_slots" to "resource_bindings" in register_tools() to match what find_tools_for_resource() expects
Wire the inferred slots into the domain Tool objects when registering MCP tools
Persist the resource bindings to the DB via the existing tool_resource_bindings migration

Metadata

Commit Message: fix(mcp): wire inferred resource slots into domain Tool objects and ToolRegistry
Branch: fix/mcp-resource-slots-wiring

Subtasks

Rename "resource_slots" to "resource_bindings" in MCPToolAdapter.register_tools() source_metadata
Wire inferred resource slots into domain Tool.resource_slots field
Persist resource bindings to DB via tool_resource_bindings table
Add unit tests verifying find_tools_for_resource() finds MCP tools with matching slots
Remove the TODO(#882) comment once fixed

Definition of Done

ToolRegistry.find_tools_for_resource("file") returns MCP tools that have a file_path parameter
MCP tool resource bindings are persisted to the DB
All existing tests pass

Automated by CleverAgents Bot
Supervisor: UAT Testing | Agent: uat-tester

## Bug Report **Feature Area:** MCP Adapter — Tool call routing / Resource slot binding **Severity:** Backlog — resource slot inference is implemented but not consumed **Found by:** UAT Testing (tool-router-mcp-adapter area) ### What Was Tested Code-level analysis of `src/cleveragents/mcp/adapter.py` — specifically the `register_tools()` method (lines 549–638) and the `infer_resource_slots()` static method (lines 652–736). ### Expected Behavior (from spec) When MCP tools are registered into the `ToolRegistry`, their inferred resource slots should be stored in the domain `Tool` objects and persisted to the database via the `tool_resource_bindings` table (migration `c1_001`). This enables the runtime to correctly bind resources to MCP tool invocations. ### Actual Behavior The `register_tools()` method infers resource slots and stores them in `source_metadata`, but there is a TODO comment acknowledging that nothing downstream reads them: ```python # TODO(#882): resource_slots are stored in source_metadata but # nothing downstream reads them yet. The ToolRegistry persists # resource bindings via domain Tool.resource_slots, and the DB # migration (c1_001) stores them in tool_resource_bindings. # A follow-up ticket should wire inferred slots into the domain # Tool objects so the registry and DB actually consume them. cap_meta = self.capability_metadata spec = ToolSpec( ... source_metadata={ "server": self._config.name, "resource_slots": slot_dicts, # Stored but never read ... }, ) ``` The `ToolRegistry.find_tools_for_resource()` method looks for `source_metadata["resource_bindings"]` (not `"resource_slots"`), so even the key name is inconsistent. ### Code Location - **Bug**: `src/cleveragents/mcp/adapter.py`, lines 611–616 (TODO comment) - **Inconsistency**: `src/cleveragents/tool/registry.py`, line 127 — looks for `"resource_bindings"` but MCP stores `"resource_slots"` ### Impact - `ToolRegistry.find_tools_for_resource()` will never find MCP tools even when they have matching resource slots - MCP tool resource bindings are never persisted to the `tool_resource_bindings` DB table - The resource slot inference logic in `infer_resource_slots()` is effectively dead code ### Fix 1. Rename the key from `"resource_slots"` to `"resource_bindings"` in `register_tools()` to match what `find_tools_for_resource()` expects 2. Wire the inferred slots into the domain `Tool` objects when registering MCP tools 3. Persist the resource bindings to the DB via the existing `tool_resource_bindings` migration ### Metadata ``` Commit Message: fix(mcp): wire inferred resource slots into domain Tool objects and ToolRegistry Branch: fix/mcp-resource-slots-wiring ``` ### Subtasks - [ ] Rename `"resource_slots"` to `"resource_bindings"` in `MCPToolAdapter.register_tools()` source_metadata - [ ] Wire inferred resource slots into domain `Tool.resource_slots` field - [ ] Persist resource bindings to DB via `tool_resource_bindings` table - [ ] Add unit tests verifying `find_tools_for_resource()` finds MCP tools with matching slots - [ ] Remove the TODO(#882) comment once fixed ### Definition of Done - `ToolRegistry.find_tools_for_resource("file")` returns MCP tools that have a `file_path` parameter - MCP tool resource bindings are persisted to the DB - All existing tests pass --- **Automated by CleverAgents Bot** Supervisor: UAT Testing | Agent: uat-tester

HAL9000 added the

labels

2026-04-09 06:57:42 +00:00

HAL9000 added this to the v3.2.0 milestone

2026-04-09 06:59:20 +00:00

HAL9000 added

and removed

labels

2026-04-09 07:00:16 +00:00

HAL9000 commented

2026-04-09 07:00:39 +00:00

Author

Owner

Label compliance fix applied:

Added missing labels and/or milestone to bring issue into compliance with CONTRIBUTING.md

Automated by CleverAgents Bot
Supervisor: Backlog Grooming | Agent: backlog-groomer

Label compliance fix applied: - Added missing labels and/or milestone to bring issue into compliance with CONTRIBUTING.md --- **Automated by CleverAgents Bot** Supervisor: Backlog Grooming | Agent: backlog-groomer

HAL9000 referenced this issue

2026-04-09 07:01:00 +00:00

[AUTO-PROJ-OWN] Project Owner Report (Cycle 26) #5296

HAL9000 added a new dependency

2026-04-09 07:02:23 +00:00

#4918 Implement concrete MCP transport layer (stdio) for tool discovery and invocation

HAL9000 referenced this issue

2026-04-09 08:20:03 +00:00

[AUTO-UAT-POOL] UAT Testing Report (Cycle 1) #5369