Tools Overview
Astonish provides 90+ built-in tools across multiple categories, plus unlimited extension via MCP servers. Tools are the agent's interface to the world — reading files, running commands, browsing the web, managing credentials, and more.
Tool Categories
| Category | Tools | Description |
|---|---|---|
| Shell & Process | 5 | Command execution, background processes |
| File & Search | 6 | Read, write, edit, search filesystem |
| Web & HTTP | 3 | Fetch pages, read PDFs, make API requests |
| Browser Automation | 34 | Full browser automation via CDP |
| 8 | Inbox management, send, search, wait | |
| Credentials | 5 | Secure secret storage and retrieval |
| Scheduler & Agent | 10+ | Scheduling, delegation, flows, planning |
| Memory | 3 | Save, search, retrieve memories |
| Drill & Testing | 7 | Create, validate, run test suites |
| Fleet | 2 | Multi-agent collaboration plans |
| Sandbox | 3 | Sandbox template management |
Confirmation System
Every tool has a confirmation level that determines whether user approval is needed before execution:
auto-approve
Safe, read-only tools that execute immediately:
read_file,file_tree,grep_search,find_filesmemory_save,memory_search,memory_getskill_lookup,list_drillsweb_fetch,read_pdf
always-confirm
Tools that modify state or have side effects:
write_file,edit_file,shell_commandhttp_request(POST/PUT/DELETE)email_send,email_replyschedule_job,distill_flow
MCP Integration
Tools from MCP servers appear alongside built-in tools. The agent sees a unified tool list and selects from both seamlessly.
MCP tools follow the same confirmation system — new MCP tools default to always-confirm until explicitly trusted.
Tool Execution in Studio
In the Studio interface, tool executions render as expandable cards showing:
- Tool name and arguments
- Execution duration
- Output (truncated with expand option)
- Confirmation buttons for
always-confirmtools - File diffs for write/edit operations
Tool Discovery
The agent can discover tools dynamically using search_tools. This is useful when a task requires capabilities not in the agent's default tool set — the agent searches for relevant tools and makes them available for the current session.
See individual tool pages for detailed documentation on each category.