GoClaw equips agents with an out-of-the-box standard library of 30+ built-in tools across 8 key categories. Agent tools are strictly governed by the global 5-layer permission framework.
Reads entire file contents within the sandbox directory scope.
Creates or replaces files with system-provided content payloads.
Applies precise line-by-line diff edits or modifications to existing files.
Recursively lists directory trees, listing file paths, sizes, and modifications.
Performs fast grepping / text searches across workspace files.
Resolves file matches based on standard wildcard pattern matching.
Spawns isolated processes to execute commands (restricted to sandboxes).
Spawns headless Chromium via Playwright to interact with pages, fill forms, and scrape data.
Searches the web for external documentation, packages, or news APIs.
Retrieves raw HTML or JSON content directly from web URLs with SSRF protection.
Searches L1 session logs and historical logs via semantic pgvector embeddings.
Retrieves precise facts and data points stored in L0 conversation indexes.
Queries the consolidated L2 semantic knowledge graph using structured searches.
Generates image files locally or via connected model pipelines.
Synthesizes audio clips or generates voice transcripts based on textual prompts.
Compiles short video files or renders graphical canvases using headless nodes.
Converts text files to natural speech MP3 structures for messaging playback.
Finds pre-configured workflow skills inside workspace repositories.
Executes complex, multi-step skills utilizing encapsulated subagent tasks.
Creates, updates, or registers new custom skill scripts on-the-fly.
Manages shared board states, creating, updating, and completing team milestones.
Launches child agents with dedicated, sandboxed subtasks and execution limits.
Hands off execution asynchronously or synchronously to another specialized team agent.
Exchanges text or files between active agent nodes on the same workspace board.
Schedules agent workflows, reports, or data syncs to run at specified intervals.
Configures agent execution loops to monitor external system health endpoints.
Queries or manages active user chat sessions, closing inactive nodes to save resources.
