WebMCP Tool Registry & Runner
All 10 tools are registered in real-time on document.modelContext and can be invoked directly by AI agents, browser extensions, or the live runner below.
WebMCP Active (Standard Runtime)10 Tools Registered
Host object: document.modelContext · Lifecycle: AbortController scoped to workspace ws_demo0001
toolchange: enabledREST proxy: 100% parity
| Tool Name | Category & Hints | Purpose | Last Execution | Action |
|---|---|---|---|---|
get_workspace_context | readread-only | Return the active workspace's id, name, goal, and a state summary (research/option/criterion counts, decision status). Call this first to orient before other tools. | Not called yet | |
search_research | readread-only | Search research items already saved in the current workspace by free-text query and optional tags. | Not called yet | |
create_research_item | write | Add a new research item (a source or piece of evidence) to the workspace's research canvas. | Not called yet | |
update_research_item | write | Modify an existing research item's title, URL, summary, tags, or evidence level. | Not called yet | |
create_option | write | Add a new comparison option (e.g. a candidate platform/tool) to the decision board. | Not called yet | |
score_option | write | Set (or update) the score and rationale for one option against one criterion. Score must be within the criterion's max score. | Not called yet | |
create_decision_draft | workflow | Generate or update the draft recommendation. If recommendedOptionId is omitted, the highest weighted-score option is auto-selected. Does not finalize anything. | Not called yet | |
request_approval | workflow | Move the current decision draft into pending_approval so a human can review and approve it in the UI. | Not called yet | |
approve_decision | consequentialhuman approval | Finalize the pending decision. This is a consequential action: it can only be executed by a human confirming in the approval dialog in the UI, never directly by an agent. Calling this as an agent will return guidance to ask the human to approve in the UI. | Not called yet | |
export_decision | readread-only | Return a structured, exportable record of the current decision: recommendation, ranking, rationale, and criteria. | Not called yet |