DeveloperWebMCP verified
WebMCP-enabled text editor with imperative tools for document I/O, workspace CRUD, on-device translation, and sub-agent (planner / researcher / writer / reviewer) delegation.
WebMCP-enabled text editor with imperative tools for document I/O, workspace CRUD, on-device translation, and sub-agent (planner / researcher / writer / reviewer) delegation. This page documents every listed tool, the action or answer it provides, its installation command, and the input fields an agent should send.
Lists all skills available in this workspace. Returns an array of { id, name, description } entries. Use read_skill to fetch the full instructions for a skill, or delegate_to_skill to run one.
Open the bandarra.me website and ask your browser agent to use list_skillsInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the list_skills tool on bandarra.me. Return the structured result and cite the source page.Returns the full definition of a skill (id, name, description, instructions, optional model) given its id or name. Use this before delegate_to_skill to inspect what a skill does.
Open the bandarra.me website and ask your browser agent to use read_skillInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the read_skill tool on bandarra.me with id: <string>, name: <string>. Return the structured result and cite the source page.id string The skill id. Either id or name must be provided.name string The skill name (case-insensitive). Either id or name must be provided.Reads the complete current editor content.
Open the bandarra.me website and ask your browser agent to use readInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the read tool on bandarra.me. Return the structured result and cite the source page.Reads the currently selected text in the editor.
Open the bandarra.me website and ask your browser agent to use read_selectionInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the read_selection tool on bandarra.me. Return the structured result and cite the source page.Finds all occurrences of a query string in the document. Returns the line and column of each match.
Open the bandarra.me website and ask your browser agent to use searchInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the search tool on bandarra.me with query: <string>. Return the structured result and cite the source page.query string The text to search for.Returns metadata about the current document: character count, word count, and line count.
Open the bandarra.me website and ask your browser agent to use get_metadataInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the get_metadata tool on bandarra.me. Return the structured result and cite the source page.Returns the current UI mode: 'editor' (Monaco editor is visible) or 'preview' (Markdown preview is visible). Check this before making edits to ensure the editor is accessible.
Open the bandarra.me website and ask your browser agent to use get_current_modeInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the get_current_mode tool on bandarra.me. Return the structured result and cite the source page.Requests the user to switch from Preview mode to Editor mode. This will display a prompt to the user and pause until they accept or decline. Call this before attempting edits when in preview mode.
Open the bandarra.me website and ask your browser agent to use request_switch_to_editorInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the request_switch_to_editor tool on bandarra.me. Return the structured result and cite the source page.Proposes a targeted edit. This tool pauses and waits for user approval. ONLY use this for small, localized changes (e.g., 1-2 sentences). Never pass the entire document.
Open the bandarra.me website and ask your browser agent to use editInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the edit tool on bandarra.me with originalText: <string>, replacementText: <string>. Return the structured result and cite the source page.originalText string The exact, minimal string of text to replace. Must be short. Do NOT pass the whole document.replacementText string The new text to replace the originalText with.Proposes a complete rewrite. This tool pauses and waits for user approval. ONLY use this when the user explicitly requests a total rewrite of the entire document.
Open the bandarra.me website and ask your browser agent to use writeInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the write tool on bandarra.me with content: <string>. Return the structured result and cite the source page.content string The full new document content.Returns the id and title of the document currently open in the editor.
Open the bandarra.me website and ask your browser agent to use get_active_doc_infoInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the get_active_doc_info tool on bandarra.me. Return the structured result and cite the source page.Lists all documents in the workspace. Returns an array of { id, title } objects.
Open the bandarra.me website and ask your browser agent to use list_workspace_docsInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the list_workspace_docs tool on bandarra.me. Return the structured result and cite the source page.Reads the full content of a specific document. Returns { title, content } or { error }.
Open the bandarra.me website and ask your browser agent to use read_workspace_docInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the read_workspace_doc tool on bandarra.me with id: <string>. Return the structured result and cite the source page.id string The document ID to read.Asks a question about a specific document using a sub-agent. Returns { summary, excerpt } where excerpt is the most relevant verbatim passage.
Open the bandarra.me website and ask your browser agent to use query_workspace_docInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the query_workspace_doc tool on bandarra.me with id: <string>, query: <string>. Return the structured result and cite the source page.id string The document ID to query.query string The question about the document.Asks a question spanning all workspace documents and synthesizes the results. Returns { summary, sources: [{ id, title, excerpt }] }.
Open the bandarra.me website and ask your browser agent to use query_workspaceInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the query_workspace tool on bandarra.me with query: <string>. Return the structured result and cite the source page.query string The question to answer across all documents.Creates a new document in the workspace with the given title and optional initial content. Providing content avoids a separate write step. Pauses for user authorization before creating.
Open the bandarra.me website and ask your browser agent to use create_documentInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the create_document tool on bandarra.me with title: <string>, content: <string>. Return the structured result and cite the source page.title string The title for the new document.content string Optional initial content for the new document. If omitted the document is created blank.Renames an existing document in the workspace. Pauses for user authorization before renaming.
Open the bandarra.me website and ask your browser agent to use rename_documentInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the rename_document tool on bandarra.me with id: <string>, title: <string>. Return the structured result and cite the source page.id string The document ID to rename.title string The new title for the document.Deletes a document from the workspace. Pauses for user authorization before deleting. If the deleted document was active, a different document becomes active.
Open the bandarra.me website and ask your browser agent to use delete_documentInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the delete_document tool on bandarra.me with id: <string>. Return the structured result and cite the source page.id string The document ID to delete.Switches the active document in the editor. Saves the current document content before switching. Does not require user authorization.
Open the bandarra.me website and ask your browser agent to use switch_active_documentInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the switch_active_document tool on bandarra.me with id: <string>. Return the structured result and cite the source page.id string The document ID to switch to.Translate a piece of text from one language to another using an on-device AI model. Languages are specified as BCP 47 tags (e.g. 'en', 'fr', 'ja').
Open the bandarra.me website and ask your browser agent to use translateInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the translate tool on bandarra.me with text: <string>, sourceLanguage: <string>, targetLanguage: <string>. Return the structured result and cite the source page.text string The text to translate.sourceLanguage string BCP 47 language tag of the source language (e.g. "en").targetLanguage string BCP 47 language tag of the target language (e.g. "fr").Delegates a task to a named skill (sub-agent). The skill runs with read-only access and returns its response as a string. Interpret the response and act on it accordingly.
Open the bandarra.me website and ask your browser agent to use delegate_to_skillInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the delegate_to_skill tool on bandarra.me with skillName: <string>, task: <string>. Return the structured result and cite the source page.skillName string The exact name of the skill to invoke.task string The specific task or instructions to pass to the skill.Delegates an ad-hoc task to a generic sub-agent. The sub-agent runs with the given system prompt and optional tool groups. Returns { result: string } with the sub-agent's final response.
Open the bandarra.me website and ask your browser agent to use invoke_agentInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the invoke_agent tool on bandarra.me with systemPrompt: <string>, task: <string>, tools: <array>. Return the structured result and cite the source page.systemPrompt string The system prompt / instructions for the sub-agent.task string The task or question to send to the sub-agent.tools array Optional tool group names to give the sub-agent. Supported: 'workspace_readonly'.Decomposes a high-level task into a structured step-by-step Plan. Returns a JSON string: { goal, steps: [{ id, instruction, dependsOn }] }. The Orchestrator reads the plan and dispatches each step using the appropriate tools.
Open the bandarra.me website and ask your browser agent to use invoke_plannerInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the invoke_planner tool on bandarra.me with task: <string>, context: <string>. Return the structured result and cite the source page.task string The high-level task to decompose into a plan.context string Optional additional context (e.g. current document summary, workspace doc list).Queries workspace documents and synthesizes a structured answer. Returns JSON: { summary, sources: [{ id, title, excerpt }] }. Use this when the task requires finding information across workspace documents before writing or reviewing.
Open the bandarra.me website and ask your browser agent to use invoke_researcherInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the invoke_researcher tool on bandarra.me with query: <string>, docIds: <array>. Return the structured result and cite the source page.query string The question or information need to research.docIds array Optional list of document IDs to restrict the search to. If omitted, all workspace documents are queried.Generates draft text for a single targeted section from an instruction and optional research/style context. Returns { draft: string } — raw text only, no edits applied. After receiving the draft, apply it using edit() for the target section. Do NOT use this to rewrite the whole document at once — use invoke_planner to break full-document tasks into per-section steps.
Open the bandarra.me website and ask your browser agent to use invoke_writerInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the invoke_writer tool on bandarra.me with instruction: <string>, researchContext: <string>, styleContext: <string>. Return the structured result and cite the source page.instruction string What to write. Be explicit: specify the target section, desired length, and any constraints.researchContext string JSON-encoded ResearchResult from invoke_researcher. Inject when the draft should cite workspace sources.styleContext string A verbatim excerpt from the document the Writer should match in tone, voice, and formatting.Evaluates a draft against explicit criteria and returns structured feedback. Returns JSON: { passed: boolean, issues: [{ severity, location?, description, fix? }], summary }. Use after invoke_writer to check a draft before applying it. If passed is false and error-severity issues remain after 3 Writer→Reviewer cycles, present the best available draft via edit() or write() and summarise remaining issues in your response.
Open the bandarra.me website and ask your browser agent to use invoke_reviewerInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the invoke_reviewer tool on bandarra.me with text: <string>, criteria: <array>. Return the structured result and cite the source page.text string The draft text to review.criteria array Review criteria to check against (e.g. 'grammatical correctness', 'consistent use of past tense', 'no unsupported factual claims').Leave a blog post comment
Open the bandarra.me website and ask your browser agent to use post_commentInstall browse first, then run the command in a terminal. WebMCP tools run directly on the website and do not need a package install.
Use the post_comment tool on bandarra.me with slug: <string>, website: <string>, author_name: <string>, author_email: <string>, text: <string>. Return the structured result and cite the source page.slug string website string author_name string author_email string text string