* Get recent commands from localStorage

Get all cached tools, grouped by server.

Resolve OpenSpace API base URL (for CLI scripts). Priority: ``cli_override`` → env var → host agent config → default.

Resolve auth headers for CLI scripts. Exits on failure.

Get platform-appropriate cache directory.

Get cursor position locally.

Resolve OpenSpace credentials and base URL. Returns: ``(auth_headers, api_base)`` — headers dict ready for HTTP requests and the

Get actual screen size from desktop environment using pyautogui. Returns: (width, height) tuple, or None on fail

Get screen size Returns: (width, height)

Get the current connection streams. Returns: The current connection (typically a tuple of read_stream, write_stream) or None if n

Lightweight summary of skills (no analyses/deps loaded). Default filters to active skills only.

Check if cache exists and has data.

* Highlight matching characters in text with <mark> tags

Shared cloud+local skill search with graceful fallback. Builds candidates, generates embeddings, runs ``SkillSearchEngine``. Cloud is attempt

Initialize the MCP session.

Dispatch by name — same routing as ShellConnector via AioHttpConnector.

Check if there is an active recording session Returns: bool: True if recording is active

Compute the Levenshtein edit distance between two strings.

List all available prompts from the MCP implementation.

Return BaseTool list. If session_name is specified, only return the tools of the specified session. If session_name is not s

List all available tools.

Load sanitized cache from disk. Returns empty dict if not exists.

With the assumption that images are screenshots that are of diminishing value as the conversation progresses, remove all but the final `image

Plan next action based on task and current screenshot. Includes prompt caching, error handling, and backup API key support.

Pre-download ALL reference files for all tasks into a local cache. Call this once before the benchmark starts so that individual task executi

Print startup banner

Read ``skills.entries.<skill_name>.env`` from OpenClaw config. This is the OpenClaw equivalent of nanobot's ``tools.mcpServers.openspace.env`

Record tool execution (internal method, called by BaseTool automatically) Args: tool_name: Name of the tool

Register a single skill directory (hot-reload). Safety: applies ``check_skill_safety`` / ``is_skill_safe`` filtering. Args:

Render entire UI

Run one phase (phase1 or phase2) over all tasks. Args: concurrency: Number of parallel OpenSpace workers. 1 = serial

Record a failed server to cache. Args: server_name: Name of the failed MCP server error: Error messa

Persist a single record (incremental — much cheaper than save_all).

Launch a coroutine as a background ``asyncio.Task``. Used by the caller (``OpenSpace._maybe_evolve_quality``) when ``background_trigg

Use an LLM to select the most relevant skills. When the local registry has more than ``PREFILTER_THRESHOLD`` skills, a **BM25 → embed

Update the tools available for evolution agent loops.

Stop the connection manager and close the connection. Args: timeout: Maximum time to wait for cleanup (default 5s).

Update legend dot colors based on theme

Update metrics