Request: MCP Authoring Usability And Chat Tool Parity

Request Summary

Close the gap between MemorySmith's runtime MCP mutation capabilities and what coding/chat agents can practically use, while enforcing prompt/runtime parity for chat tool usage.

Why This Request Exists

During the 2026-05-29 MCP task backlog pass, task records were created by writing JSON files directly instead of using MemorySmith MCP task tools. That was not because the product lacks task/page mutation tools. The runtime already exposes MCP task and page mutation tools, but this VS Code coding-agent environment does not have those repo-local MemorySmith MCP tools wired in as callable tools.

At the same time, the chat prompt intentionally allows only a smaller intercepted tool subset and routes memory/page writes through proposal JSON. That is valid policy, but the split between:

• runtime MCP tool descriptors,
• chat prompt tool lists,
• agent-mode mutation allowances,
• and external coding-agent tool availability

is large enough that parity and usability drift are easy to introduce.

Evidence

• README.md documents MCP mutation tools including memorysmith_task_create, memorysmith_task_update, memorysmith_task_set_status, memorysmith_task_add_comment, memorysmith_page_save, and memorysmith_page_delete.
• MemorySmith.App/Services/ChatToolCatalog.cs exposes those task/page mutation tools with AvailableInMcp: true, while chat availability stays off for mutation operations.
• MemorySmith.App/Controllers/McpController.cs confirms MCP calls can execute any tool marked AvailableInMcp and enabled by config.
• MemorySmith.Core/Docs/Prompts/wiki-chat-agent.md limits intercepted chat tools to read-oriented tools and states that task mutation tools are agent-only and memory/page changes must flow through proposal JSON.
• This coding-agent session had no callable MemorySmith MCP task/page tools, so direct file authoring was the only executable path for creating task records from here.

Recommendations

1. Add MCP-backed authoring helpers for agent workflows

• Classification: Now
• Impact: high
• Effort: medium
• Confidence: 94%
• Recommendation: provide a first-class bridge so coding agents can create/update tasks and draft page writes through MemorySmith MCP without hand-authoring JSON files.
• Preferred shape:
• expose the local MemorySmith MCP server to the coding-agent environment as callable tools, or
• add a thin repo script/helper that submits validated MCP tools/call requests to the local app for task/page authoring.

2. Make chat tool usage parity a first-class contract

• Classification: Now
• Impact: high
• Effort: medium
• Confidence: 92%
• Recommendation: define and test explicit parity rules across README.md, wiki-chat-agent.md, ChatToolCatalog.cs, and any agent capability surfaces so tool availability differences are intentional, documented, and verifiable.
• Minimum parity goal: perfect agreement on what Chat mode, Agent mode, MCP mode, and coding-agent integrations can and cannot call.

3. Add authoring-oriented MCP ergonomics

• Classification: Next
• Impact: medium-high
• Effort: medium
• Confidence: 89%
• Recommendation: add dryRun/validateOnly, task-template defaults, id/key allocation hints, and structured envelopes for authoring results so agents can create artifacts without building full JSON by hand.

4. Add parity-audit automation

• Classification: Next
• Impact: medium-high
• Effort: low-medium
• Confidence: 90%
• Recommendation: add a focused parity verifier that compares documented tool sets and mode flags against the runtime catalog, with special attention to chat tool usage.

5. Keep proposal-first memory/page mutation policy explicit

• Classification: Later
• Impact: medium
• Effort: low
• Confidence: 88%
• Recommendation: retain the policy distinction where direct task mutation may be agent-capable but memory/page mutation remains proposal-first unless governance explicitly widens it.

Status

InProgress

Active tracking:

• TSK-0224 (primary parity contract task)

Follow-Up

• Track execution in TSK-0223 and TSK-0224.
• Fold any resulting implementation into the prompt/runtime parity audit cadence and MCP recommendation backlog.

2026-05-30 Self-Review Ordering

• Now (confidence 92%): keep parity requirements explicit across prompt/runtime/docs and continue TSK-0224 as the contract owner.
• Next (confidence 90%): add scriptable parity drift checks that compare runtime tool catalog mode flags against documented/intercepted tool sets.
• Later (confidence 88%): add coding-agent surface conformance checks once MCP bridge ergonomics from TSK-0223 stabilize.