Built-in Tools
Enable and configure the tool capabilities that ship with Clouisle
Built-in tools are registered by the platform and never created or deleted. Select one under Capabilities > Tools to test it; tools that need credentials show "Configuration needed", and Configure stores the team credentials. Tool details list the parameter name, type, required flag, enum, and default value.
Selectable Built-in Tools
These tools appear in the tool catalog and can be bound to an Agent directly:
| Tool ID | Purpose | Extra configuration |
|---|---|---|
get_current_time | Current date, time, weekday, and timestamp for a time zone (timezone_name, default UTC) | None |
format_datetime | Format a timestamp with strftime (timestamp defaults to now; format_string defaults to %Y-%m-%d %H:%M:%S) | None |
calculate | Math calculation (expression) | None |
unit_convert | Unit conversion (value, from_unit, to_unit) | None |
web_search | Web search (query required; num_results default 5, max 10) | TAVILY_API_KEY / BOCHA_API_KEY (optional) |
fetch_webpage | Fetch a page and extract its text (url required; max_length default 5000 characters) | None |
rss_feed_reader | Parse an RSS / Atom feed (url required; limit default 10, max 30) | None |
markitdown | Convert uploaded PDF / Word / Excel / PPT / text files to text (files_url required, an array of URLs) | None |
artifact / bash / edit / read / write | Artifact management and file I/O inside the code sandbox | Sandbox runtime |
Search backends for web_search
web_search defaults to the auto engine: it prefers a configured commercial backend — Tavily when TAVILY_API_KEY is set, otherwise Bocha when BOCHA_API_KEY is set — and falls back to key-free DuckDuckGo when neither is configured. You can also force tavily / bocha / duckduckgo; in auto mode a failing commercial backend automatically falls back to DuckDuckGo. Credentials live in server-side team configuration — never write them into an Agent prompt.
Tools injected by capability switches
The following tools are not in the catalog and cannot be attached through tools_config; each is injected by its own Agent capability switch:
| Tool ID | Switch | Description |
|---|---|---|
ask_user | User input (enable_user_input_request) | Asks the user an interactive question form and waits for the answer |
generate_image | Image generation (enable_image_generation) | Generates images with an authorized image model |
generate_video | Video generation (enable_video_generation) | Generates videos with an authorized video model |
Memory tools
When Agent memory is enabled (enable_memory), these 5 memory tools are injected. auto_extract only gates writes — the read tools stay available when it is off:
| Tool | Purpose | Limits |
|---|---|---|
search_memory | Find relevant memory entities | query required; top_k defaults to 5 and is clamped to 1-10; supports time_window_days and entity_type filters |
get_memory_subgraph | Read related entities and relations | entity_ids at most 5 (entity names or IDs); max_depth 1-3, default 1; direction is incoming / outgoing / both (default both) |
create_memory_entity | Create a memory entity | Write tool; unavailable when auto_extract is off |
create_memory_relation | Create a relation between entities | Write tool; unavailable when auto_extract is off |
update_memory_entity | Update an entity's description or properties | Write tool; unavailable when auto_extract is off |
Use search_memory to find candidate entities first; call get_memory_subgraph only when the answer depends on relationships. See Memory and Context.
Execution results and artifacts of the sandbox tools (bash / read / edit / write / artifact) must be accessed per team permissions; never share internal download URLs.
Built-in tools cannot be deleted or shared with other teams; only custom tools support enable/disable and cross-team sharing.
How is this guide?