Loading ToolBench
AI workflow engine: turns your project docs into a structured graph your AI agent reads and writes. Provides tools for managing weaves, threads, ideas, designs, plans, requirements, and chats.
Loom MCP has 55 tools with minimal schema visibility and sparse descriptions. Source code inspection reveals tool files exist but actual parameter schemas, input/output types, and detailed descriptions are not visible in the provided code samples. Tool names follow verb_noun convention (create_*, set_*, update_*) which is positive, but without access to the actual tool registration code in packages/mcp/src/tools/*.ts, schemas cannot be verified. Descriptions provided in the tool list are 1-2 sentences (10-50 chars), well below the 50-200 char baseline for LLM-optimized descriptions. No evidence of error handling patterns, output schema documentation, or parameter validation guidance. STDIO-only transport caps protocol readiness at 50.
Add a new step to a plan
Amend an existing requirements document
Append to the done document
Append a message to a chat document
Archive a document or thread
Close a plan
Mark a plan step as complete
Create a new chat document
Create a design document in an existing thread
No input schemas visible for any of 55 tools. Cannot verify parameter types, required fields, constraints, or descriptions. All tools score 0 on schema dimension.
Tool descriptions are 1-2 sentences (10-50 chars), far below the 50-200 char baseline for LLM-optimized descriptions. Descriptions lack WHAT/WHEN/WHY context needed for tool selection.
No evidence of error handling patterns. Tools like loom_delete (DESTRUCTIVE) and loom_finalize_req (WRITE) lack confirmation/dry-run support or recovery guidance.
Tool names like 'loom_do_step', 'loom_quick_ship', 'loom_refresh_ctx' are vague. 'do_step' does not convey what action occurs; 'quick_ship' is colloquial; 'refresh_ctx' is unclear without domain knowledge.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 31 | 2026-07-28+ | v2 |
Create an idea document in an existing thread
Create a plan document in an existing thread
Create a reference document
Create a report document
Create a requirements document in an existing thread
Create a new thread in a weave
Create a new weave
Delete a document or thread
Execute a plan step
Finalize (lock) a requirements document
Find a document by ID or slug
Generate content for documents using AI
Get plan steps that are blocked by dependencies
Get context preferences for documents
Get documents that are stale (not updated recently)
Get plans that are stale (not updated recently)
Install Loom into a workspace
List steps in a plan
Move a thread to a different weave
Patch a document with targeted changes
Promote a document to a higher type (e.g., idea to design)
Quick ship a plan (complete and close)
Read the tail (recent messages) of a chat document
Record a release for shipped plans
Refine a design document
Refine an idea document
Refine a plan document
Refresh the global context document
Remove a step from a plan
Rename a document or thread
Rename a document file
Rename a thread
Rename a weave
Reorder steps in a plan
Restore an archived document or thread
Search documents by title or content
Seed an example Loom workspace
Set context preferences for documents
Set the priority of a thread or weave
Set the status of a document or thread
Set dependencies for a thread
Start executing a plan
Update a document with new content
Update a plan step
Validate document integrity and structure
Verify a thread against its requirements
No output schema documentation visible. LLMs cannot plan downstream tool calls or extract required fields (e.g., thread_id, weave_id) for chaining.