Open Claw Runtime Tools

Diagram of the Open Claw runtime tools, the System Prompt and the Open Claw Tool Defintions

Runtime Capabilities

• Memory: recall prior facts and context.
• Sandbox: safely read, write, edit, and execute code.
• Cron: run jobs on a schedule.
• Skills: packaged workflows and constraints.
• Toolsets (OpenAPI, web, etc.): connect external systems.

Without a runtime, you have a chat model. With a runtime, you have an agent that can act reliably over time.

Tool Summary

Tool name	Category	Required params	Key enums / notes
read	Filesystem	path | file_path	Read text or images, supports offset, limit
write	Filesystem	content, path | file_path	Overwrites file
edit	Filesystem	path | file_path, old*, new*	Exact string replace; multiple alias params (oldText, old_string)
exec	Shell	cmd	Long-running allowed; timeout, pty, background, elevated, host, security
process	Shell	action	logs, write, keys, kill, status (exec session control)
browser	Browser automation	action	Large dispatcher: start, stop, open, navigate, act, snapshot, pdf, etc.
canvas	UI / A2UI	action	present, hide, eval, snapshot, push, reset
nodes	Device / node control	action	Pairing, notify, camera/screen/location, run, invoke
message	Messaging	action, content	Only send despite description mentioning broadcast
tts	Audio	text	Text-to-speech
agents_list	Agent mgmt	—	Lists available agents
sessions_list	Session mgmt	—	List sessions
sessions_history	Session mgmt	session_id	Fetch conversation history
sessions_send	Session mgmt	session_id, content	Send message to session
sessions_spawn	Session mgmt	content	Spawn sub-agent (one-shot or persistent)
subagents	Agent mgmt	action	list, kill, send
session_status	Session mgmt	session_id	Inspect session; optional model override
web_search	Web	query	Brave Search wrapper; locale options
web_fetch	Web	url	Fetch + readable extraction
memory_search	Memory	query	Semantic search over memory
memory_get	Memory	id	Retrieve memory entry