Browse

List all Google AI cached content entries. Returns a JSON array of cache names. Only available when GOOGLE_API_KEY is set.

List the GGUF files in one HuggingFace repo with their sizes, to pick one for /model hff download.

Search HuggingFace for GGUF model repos (most downloaded first) and list each repo's GGUF files with their sizes. Use to find a local model to download with /model hff download.

Make an HTTP request to a URL. Returns response status, headers, and body. Use for calling APIs, fetching web content, or interacting with external services. Requires: url. Optional: method (default GET), headers, data (JSON body), files (multipart uploads), timeout, failOnError, saveTo, retry, cache, connectionName, accountName, followRedirects, tls.

List files and directories in a given directory path. Returns names and types (file/dir). Use to discover project structure before reading or editing files. Omit path to list the current working directory.

Compute the MD5 checksum of a file. Call once before and once after a change to prove whether the file's contents actually changed -- pair both hashes as evidence rather than asserting the change happened.

Extract text from an image via tesseract OCR. Runs entirely locally -- no API key required. Supports png, jpg/jpeg, tiff, bmp, gif. Requires: file (absolute path to the image). Optional: language (tesseract -l value, e.g. 'eng' or 'eng+fra'; default 'eng'), psm (page segmentation mode, default 3), oem (engine mode, default 3). Does not support PDF -- use load_document with type pdf for text-layer PDFs.

kdeps tool: perplexity_search

Run a Python script and return its result: stdout parsed as JSON when possible, otherwise the raw text. Use through the python component for pipeline steps written in Python. Give either script (inline source) or scriptFile (a .py file). Optional: args, env, timeout, venvName, pythonVersion, packages.

Read a file from the local filesystem. Returns the contents with a 1-based line number on every line (` 42\tcode$`) - use those numbers for edit_file insert/view. Tabs render as ^I, other control characters as ^X, and a $ marks the true end of each line, so indentation, trailing whitespace, and stray control characters are visible instead of hidden. IMPORTANT: ^I, ^X, and the trailing $ are a display convention this tool adds - the real file does NOT contain those literal characters. When quoting text back into edit_file's old_str, write the real tab/control character (or match a whole line via view_range/anchor instead), never the literal caret sequence - copying "^I" as text will not match a real tab byte. Plain text, source code, configuration files, and documentation are read directly. For PDF, DOCX, EPUB, RTF, and ODT documents, CSV/HTML structured parsing, or RAG chunking, use load_document. Pass match_id (from a search_local result) instead of file_path/offset to jump straight to that hit.

kdeps tool: serpapi_search

Read or change the active agent session: messages lists every message ({turns, messages: [{turn, role, content}]}); last_reply is the latest assistant message ({text}); get returns a session setting's state ({value, ...}); set stores a value the setting's command asset lists and returns the new state; clear empties the history, summarizing it first when it has at least summarize_min_turns turns ({turns, summary}); compact folds older turns into a summary ({summary, turns, keep}); sessions lists this folder's saved sessions ({sessions: [{id, name, prompt, model, turns, updated}]}); save stores the session ({id, name}); load and import make a saved session (or a JSONL file at path) the live one ({id, turns}); delete removes one ({id}); checkpoint marks the current entry ({id}); goto returns to an entry, stashing the turns after it ({id, found, turns}); branches lists the stashed ones ({branches: [{branch, at, ids}]}); model switches the model (name: a model, a .gguf/.llamafile URL, or an OpenAI-compatible base URL; {model, registered?, ambiguous?, closest?}); model_pick opens the model picker ({model, picked}); model_default saves the startup model ({model}); models lists cloud models by backend and local ones ({current, cloud, local}); model_favorite stars a model (on: on|off); loop_settings reads the agent loop settings and loop_set changes one (setting, value); servers lists running local model servers and server_kill stops one (model); hf_download downloads a GGUF file from HuggingFace and registers it (repo, file; {path, alias}); model_record is the current model's record of caught refusals ({model, count, refusals: [{kind, text, at}]}) and model_record_clear deletes it ({model, cleared}); skills and prompts list the loaded skills ({skills: [{name, description}]}) and prompt templates ({prompts: [{name, hint, description}]}); reload rereads them from disk ({skills, prompts}: counts); memory_entries lists memory entries, all or the matches of query, by key or (order recent) newest first, capped by limit ({total, more, query, entries: [{key, type, value}]}); memory_entry is one entry in full ({key, type, value, related, updated, graph}); judges is the judge panel ({judges: [{name, criteria, graders}], auto, last}); judge_add (name, criteria), judge_remove (name), judge_grade (name, kind, pattern) and judges_clear change it; harness_sections, harness_events, presets and harness_reminders list the harness sections, events, presets and reminders, and harness_set, event_set, preset_apply and reminder_set change them (name, enabled: on|off); efficiency is enforcement's state ({value, verbose, prompt, limits}), efficiency_limit sets one limit (setting, value: a count or off) and efficiency_reset restores the defaults; recall recalls knowledge vault notes for query, or returns the last turn's recall ({seeds, hits, tokens}); knowledge is the vault's state ({value, vault, graph}) and knowledge_extract reads the changed notes with the local model ({output}); last_request is the last request sent to the LLM with token estimates (raw: true for JSON); search_index, search_folders and search_folder (action: add|remove, folder) manage the folders the project searches; commands_help lists every installed command's usages; konfig_export and konfig_import write and apply a konfig file (path); pipelines is /kartographer's live state; tool runs one tool that exists only in the agent session (judge_verdict, knowledge_recall, knowledge_card_get, knowledge_card_update, knowledge_note_save, load_skill, identity_get, session_handshake) with args and returns its answer.

Return the schema (column names and types) for a table in a SQLite database. Use before writing queries to know the exact column names.

List all tables in a SQLite database. Use this to discover available data before querying. Returns a newline-separated list of table names.

Run SQL against a named connection in config.yaml sql_connections and return the rows or a change summary. Databases: PostgreSQL, MySQL/MariaDB, SQLite, SQL Server, Oracle. One statement (query + params), or several in one transaction (transaction + queries). Requires: connectionName.

Execute a SQL query against a SQLite database and return the results as formatted text. Use SELECT statements to retrieve data. Non-SELECT statements are rejected for safety.

Read the last N lines of a file (default 20) -- for checking the end of a log or output file without knowing its total length upfront. Use as evidence that a command's output or a log ends the way a task claims it does.

Transcribe an audio or video file to text through a Whisper API: openai (default), groq, or local (a self-hosted OpenAI-compatible Whisper server). Supports mp3, mp4, mpeg, mpga, m4a, wav, webm. Returns the transcribed text. For offline transcription with no API key use transcribe_audio_offline.

Transcribe an audio file to text offline with whisper.cpp (whisper-cli): no API key, no network. Supports flac, mp3, ogg, wav. Uses the GGML model at modelPath, or ggml-base.en.bin in the kdeps models dir. The model is not downloaded: place the GGML file there first. Returns the transcribed text.

kdeps tool: twiml

Run one vector store operation against an external vector database: add_documents to upsert documents, or similarity_search to query them. Call it through the vector_store component. Requires: collection, operation, embedModel (except bedrock). Optional: provider, url, apiKey, documents, query, topK, embedBackend, embedBaseURL.

Fetch and extract readable text content from any web URL. Returns page title, headers, body content, and links. Use when you need to read a specific web page, article, or documentation URL.

Search the web using DuckDuckGo. Free, no API key required. Use for current events, facts, research, or anything needing an internet lookup. Input is a plain search query string.

Look up information on Wikipedia. Use for general knowledge questions about people, places, companies, historical events, concepts, or any topic needing an encyclopedic answer. Input is a search query.

Query Wolfram Alpha for factual computations, math, science, unit conversions, and data lookups. Returns a concise plain-text answer. Requires WOLFRAM_APP_ID.

Write or overwrite a text file on the local filesystem. Creates a new file if it does not exist; overwrites existing files entirely. Use for creating or updating configuration files, source code, scripts, or any text-based file. Takes an absolute path.

List all available Zapier NLA actions configured in your Zapier account. Returns action IDs, names, and descriptions. Use this to discover what actions you can run with zapier_run_action. Requires ZAPIER_NLA_API_KEY.

Execute a Zapier NLA action by action ID with natural language instructions. First use zapier_list_actions to find available action IDs. Requires ZAPIER_NLA_API_KEY.