Loading ToolBench

Reading the current corpus…

core MCP server · B · ToolBench

core

Server-hosted MCP-over-HTTP resource providing agent tools for conversation management, expert routing, knowledge curation, and project administration. Implements OAuth-gated access with mcp_at_ tokens and in-process execution against the core Orako application layer.

<=2025-11-25STREAMABLE_HTTP12 tools

70/100

Definition Quality81

Protocol Readiness72

Supportability51

Spec posture

Inferred effective spec <=2025-11-25·current baseline 2026-07-28

Static source inference · medium confidence · detected: Logging, Dynamic Client Registration

Deprecated protocol patterns detected

Summary

Orako MCP server demonstrates solid tool design with consistent naming (verb_noun pattern), comprehensive parameter descriptions, and well-structured schemas. All 12 tools have clear descriptions (avg ~150 chars) and typed input parameters. Tool composition is thoughtful, search_history, list_experts, ask, follow_up, resolve_conversation form a coherent conversation-management workflow. However, output schemas are not explicitly documented in the source code provided, limiting visibility into response structure. Error handling guidance is absent, tools lack recovery hints or actionable error messages. Some parameter descriptions could be more prescriptive about constraints (e.g., top_k bounds, tag/entity normalization rules). Security considerations (permission gates, audit trails) are not visible in the tool definitions themselves.

Tools (12)

add_knowledgewriteauthsource verified82/100

Create a curated knowledge entry (FAQ/runbook answer). Admin-gated tool for team-authored knowledge base entries. Entry is immediately searchable and attributed to the organization.

add_participantwriteauthsource verified80/100

Add an additional expert to an open conversation thread. Allows agent to expand the pool of responders mid-conversation.

askwriteauthsource verified83/100

Open a new conversation by asking one expert (direct ask via member_id) or dispatching to a pool of experts matching specified domains. Requires a curated one-line summary (powers history search) and optional title. Returns conversation ID.

follow_upwriteauthsource verified82/100

Add a follow-up message to an open conversation thread. Allows agent to continue discussion, ask clarifying questions, or provide additional context.

get_conversationread onlyauthsource verified78/100

Fetch full conversation details including all messages, participants, and metadata. Used to read thread state before follow-up or resolution.

list_conversationsread onlyauth

Issues (5)

HIGHtool ↗

Output schemas not documented in source. Tools return data but LLMs cannot see field structure, types, or required fields. Breaks downstream tool chaining and forces agents to infer response shape.

search_historylist_expertsaskget_conversationlist_conversations

HIGHrecovery-guide ↗

Error handling lacks recovery guidance. Tools have no documented error cases, retryability classification, or actionable error messages. Agents cannot self-correct on failure.

askfollow_upresolve_conversationupload_attachment

MEDIUMconstrained-input ↗

Parameter constraints underspecified. 'top_k' (search_history) lacks min/max bounds. 'tags' and 'entities' (ask, resolve_conversation) lack normalization rules or enum examples. LLMs may pass invalid values.

search_historyaskresolve_conversationadd_knowledgesuggest_knowledge

Recommendations

Document output schemas for all tools. For search_history, specify: {conversations: [{id, title, status, asker, age, matched_snippet}], total_count, next_cursor}. For list_experts, specify: {members: [{id, name, domains, delivery_channels}]}. This enables LLM planning and response parsing.
Add error handling guidance to write-heavy tools. Example for 'ask': 'Returns {conversation_id} on success. Errors: invalid_project_id (404, try list_projects first), invalid_member_id (404, try list_experts), invalid_domains (400, must match known expertise domains). Retryable: false for 400/404, true for 5xx.'
Specify parameter constraints explicitly. For search_history.top_k, add '(integer, 1 - 100, default 20)'. For ask.tags/entities, add '(array of strings, normalized lowercase, max 10 each, reuse known_tags/known_entities from search_history response)'. For upload_attachment.data, add '(base64-encoded, max 50MB)'.
Document permission requirements in tool descriptions. For add_knowledge, add: 'Requires admin:knowledge scope. Non-admins should use suggest_knowledge instead.' For add_participant, add: 'Requires write:conversation scope and membership in target conversation's project.'
Add pagination to list tools. For list_conversations, add parameters: limit (1 - 100, default 20), offset (integer, default 0). Return {conversations: [...], total_count, has_more}. For search_history, add cursor parameter and document result capping at 50 matches.

Spec posture evidence

Inferred effective spec: <=2025-11-25.

Relies on Logging (deprecated) - log to stderr or use OpenTelemetry

Score history

Overall score trend

First recorded score · v2 rubric

70/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	B	70	<=2025-11-25	v2