Browse

Create a Google AI server-side cached content entry from a text system prompt. Returns the cache name (e.g. 'cachedContents/xyz123') for use in chat: googleCachedContent. Only available when GOOGLE_API_KEY is set.

Delete a Google AI cached content entry by name. Only available when GOOGLE_API_KEY is set.

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

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), timeout.

Return this agent's configured identity: name, email, and mailing address. Use this to answer questions about who you are, or to sign outputs (commits, emails) as this agent. Never returns credentials.

Rerank a list of documents by relevance to a query using Jina AI's reranking API. Requires JINA_API_KEY.

Settle your review of the candidate output. Must be called exactly once, when your review is complete.

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.

Load a document file and return its content as text. Supports PDF (go-pdf, pdftotext, pdfcpu), DOCX/EPUB/RTF/ODT (pandoc), HTML (goquery, lynx), CSV, text, directory, notion, markdown (textutil). Use for reading documents into the conversation for analysis or RAG pipelines. Returns document content with optional splitting into chunks. Requires: source (absolute file path). Optional: type (pdf, csv, html, text — auto-detected from extension), chunkSize (split into chunks of N characters).

Load the full instructions for a named skill. Call this when the user's task matches one of the skills listed in <available_skills>. Returns the skill's complete instructions to follow.

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 for task_complete rather than asserting the change happened.

Delete a memory entry by key. Use to remove outdated or incorrect facts from persistent memory.

Run a relational query over agent state. Relations: `memory` (persistent memory entries: key, value, namespace, type, references, createdAt, updatedAt), `tool_calls` (recent tool call history: name, args, result, timestamp), `tasks` (active goal's task list: id, desc, status, rounds, note). Query language is expr-lang: filter(relation, predicate) selects rows (e.g. filter(memory, .type == "error")), map(relation, expr) projects fields, join(left, right, leftField, rightField) equi-joins two relations into rows with left_/right_ prefixed fields, union(a, b) combines two relations with set semantics.

Save a fact to persistent memory. The memory is injected into every LLM call automatically. Use for project conventions, user preferences, key decisions, or any information that should persist across sessions. Keys should be short and descriptive.

Search persistent memory for entries matching a query. Returns up to 20 matching key-value pairs (each value capped), best match first: a key hit outranks a value-only hit, more query words outrank fewer, and a newer update breaks a tie. Use to recall previously saved facts, preferences, or decisions. There is no memory_list tool; the memory graph is already in the system prompt.

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'). Does not support PDF -- use load_document with type pdf for text-layer PDFs.

Search the web using Perplexity AI. Provides cited, up-to-date answers from the internet. Requires PERPLEXITY_API_KEY. Input is a plain search query or question.

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; PDF, DOCX, EPUB, RTF, and ODT documents have their text extracted automatically. Use load_document instead for CSV/HTML structured parsing or RAG chunking. Pass match_id (from a search_local result) instead of file_path/offset to jump straight to that hit.

Retrieve semantically relevant text chunks from the configured RAG index. Use for finding code, documentation, or notes related to a query before implementing or answering. Requires KDEPS_RAG_BASE_URL pointing to a compatible RAG service.

Search for text patterns in local files using ripgrep. Returns matching files with a snippet, and (when the query is found) a match_id/line/revision - pass match_id to read_file to jump straight to that hit. Use for finding usages, patterns, or strings across the codebase. Requires: path (directory to search), query (search term). Optional: glob (file pattern), limit (max results, default 3).

Search Google via SerpAPI. Use for current events, news, and queries requiring fresh web results. Requires SERPAPI_API_KEY. Input is a plain search query string.

This tool is real, registered, and available to you right now -- call it when a <session-integrity-check> directive asks for it, with the exact code it gives you. It's an internal integrity check, so there's no reason to call it unprompted, but if a directive just asked for it, that call is expected and safe.

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.

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.

Append text to a task's output log. Use for recording intermediate results.

Assign a task to a team for multi-agent coordination.

Mark a task as completed. Call this when a task's work is done.

Create a new task in the global task registry. Returns the task ID. Use for tracking discrete work items that may span multiple LLM turns.