Index, search, and prepare local Codex sessions for Hermes and MCP clients, with OpenClaw compatibility, approval-gated dry-runs, and optional Codex controls.
Critical failure across all 53 tools. ZERO input schemas visible in source code, all tools declare Input: {} with no parameter definitions. Descriptions are present but generic and lack actionable detail (avg ~60 chars, well below 194-char baseline). No parameter descriptions, no output schemas documented, no error handling guidance. Tools are inferred from registration lists rather than explicit schema definitions. This server is non-functional for LLM integration.
Review the compact attention queue before choosing a next action.
Tail audit logs
Get business pulse information
Perform a dry-run closeout operation
Get Codex app server status
Get Codex app server threads
Perform a dry-run Codex control operation
ALL 53 TOOLS: Input schema is empty object {} with zero parameter definitions. No types, no descriptions, no constraints. LLMs cannot determine what parameters to pass or what values are valid.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 32 | 2025-06-18+ | v2 |
Resolve one daemon-owned Codex target or return the explicit Desktop observation route without exposing raw identifiers.
Deliver through the opaque target using idle send or active steer under the existing dry-run approval audit.
Extract Codex information
Interrupt an active Codex thread
Resume an existing Codex thread
Send a message to a Codex thread
Access Codex SQLite stores
Start a new Codex thread
Steer an active Codex thread
Look up a specific session or source ref after the inbox identifies it.
Perform desktop GUI actions
Prove desktop visibility and capabilities
Prove desktop action execution
Diagnose LCO setup and health
Drive Codex operations
Expand one bounded evidence brief from a query when the ref is not known.
Expand session details
Find existing indexed local work from one query with public-safe refs; refresh first with lco_index_sessions or index:true when needed.
Grep through session data
Index local Codex sessions
Access LCM peer databases
Get operating picture of sessions
Check and manage permissions
Start from the compact prepared-state operating picture.
Get prepared state information
Create a bounded provenance and handoff digest from available operating inputs.
Refresh recent or active session cards after reads or approved actions.
Search indexed Codex sessions
Inspect what changed since a previous opaque cursor before drive/control planning.
Sanitize session data
Get visible Codex session map
Manage watchers for session monitoring
Alias for lco_codex_extract with kind:final_messages
Alias for lco_codex_extract with kind:plans
Alias for lco_codex_extract with kind:tool_calls
Alias for lco_codex_extract with kind:touched_files
Alias for lco_describe_ref
Alias for lco_prepared_state with view:cards
Alias for lco_prepared_state with view:status
Alias for lco_watchers with action:resume_request_packet
Alias for lco_prepared_state with view:expand
Alias for lco_prepared_state with view:leaves
Alias for lco_watchers with action:dry_run
Alias for lco_watchers with action:events
Alias for lco_watchers with action:status
Alias for lco_watchers with action:list
No output schemas documented. LLMs cannot predict what fields will be returned or plan downstream tool calls. No pagination info, no field descriptions, no structured response guidance.
Alias tools (loo_* prefix, 13 tools) have descriptions under 20 characters ('Alias for lco_X with action:Y'). These are inferred aliases, not explicit tool registrations, cap overall at 50.
No error handling guidance. Tools like lco_drive and lco_codex_deliver (WRITE operations) have no documented error recovery, retry logic, or confirmation steps. Agents cannot distinguish retryable from fatal failures.
Tool descriptions are generic and lack actionable context. E.g., 'Manage watchers for session monitoring' does not explain WHEN to call it, WHAT parameters it accepts, or WHAT it returns. Average description length ~60 chars vs. 194-char baseline.