Skip to content

Built-in Tools

Wireclaw agents have access to 30+ built-in tools across 8 categories. Enable them in your agent config:

[tools]
enabled = ["web_search", "browser", "shell", "file_read", "file_write"]
ToolDescription
web_searchSearch the web via DuckDuckGo or Brave. Returns titles, URLs, and snippets.
web_fetchFetch a URL and convert HTML to Markdown. Useful for reading articles, docs, pages.
http_requestMake raw HTTP requests (GET, POST, PUT, DELETE). For APIs, webhooks, external services.
[tools.web_fetch]
domain_allowlist = ["docs.example.com", "*.github.com"]
[tools.http_request]
domain_allowlist = ["api.example.com"]
ToolDescription
browserFull browser automation via Playwright — screenshot, click, fill, navigate, evaluate JavaScript.

The browser tool gives agents a real browser for interacting with dynamic web pages, filling forms, and capturing screenshots.

[tools.browser]
domain_allowlist = ["*.wikipedia.org", "arxiv.org"]
ToolDescription
file_readRead files from the agent’s workspace.
file_writeWrite or create files in the workspace.
file_editEdit existing files with search-and-replace operations.
glob_searchFind files by glob pattern (e.g., **/*.md).
content_searchSearch file contents with regex patterns.

File operations are scoped to the agent’s workspace directory — agents cannot access files outside their workspace.

ToolDescription
shellExecute shell commands in a sandboxed environment.
[tools.shell]
sandbox = "landlock" # Options: landlock, bubblewrap, firejail, docker
timeout_secs = 30
domain_allowlist = ["api.github.com"] # Network access restrictions

Sandbox backends:

BackendOSIsolation
landlockLinux 5.13+Filesystem access control
bubblewrapLinuxContainer-like isolation
firejailLinuxSecurity sandbox
dockerAnyFull container isolation
ToolDescription
memory_storeStore a key-value pair in persistent memory.
memory_recallRetrieve stored memory by key or semantic search.
memory_forgetDelete a memory entry.

Memory persists across sessions and restarts. See the [memory] section in Agent Config.

ToolDescription
cronCreate, list, update, remove, and run scheduled tasks.
scheduleSchedule a one-time future action.

Agents can schedule recurring tasks (e.g., “check this URL every hour”) or delayed actions.

ToolDescription
pdf_readExtract text and structure from PDF files.
ToolDescription
delegateCall another agent to handle a subtask.
pushoverSend push notifications via Pushover.

Control how much oversight tools require:

LevelBehavior
autonomousTools execute without approval (default).
guidedTools execute but all actions are logged.
supervisedTools require explicit user approval before execution.
[tools]
autonomy = "supervised" # Require approval for all tool actions