# Tool reference > Every tool the connector exposes, grouped by what it touches, with its display title and whether it reads or writes. Source: https://docs.generalinput.com/mcp/tools The toolset is the platform's agent toolset, single-sourced, so it is identical to what a General Input chat can do. Display titles are what most hosts show; the snake\_case name is what the model calls. Tools not marked read-only are writes, and hosts that gate on annotations ask before running them. ## Integrations and credentials [#integrations-and-credentials] | Tool | Title | Reads only | Does | | --------------------------------- | ------------------------------- | ---------- | ---------------------------------------------------------------------------------- | | `search_internal_api_docs` | Search Integration Actions | yes | Hybrid search across every operation the workspace can use. | | `get_internal_api_operation_docs` | View Integration Action Details | yes | One operation's full reference. | | `search_connected_credentials` | View Connected Accounts | yes | Credentials the member can use, with env var names. | | `user_connect_credential` | Ask to Connect an Account | | Returns a dashboard link to connect a service. | | `bash` | Run a Command | | Run a shell command in the sandbox with declared credentials injected. Open-world. | ## Sandbox and files [#sandbox-and-files] | Tool | Title | Reads only | Does | | ------------------- | ------------- | ---------- | ----------------------------------------------------------- | | `read_file` | Read a File | yes | | | `write_file` | Write a File | | | | `edit_file` | Edit a File | | Targeted string replacement. | | `delete_file` | Delete a File | | Destructive. | | `list_files` | List Files | yes | | | `search_files` | Search Files | yes | | | `send_file_to_user` | Send a File | | Upload a sandbox file and return a seven-day download link. | ## Web [#web] | Tool | Title | Reads only | Does | | ----------------- | --------------- | ---------- | -------------------------------------------------------- | | `search_internet` | Search the Web | yes | Open-world web search. | | `scrape_webpage` | Read a Web Page | yes | Fetch a page through the proxy scraper. Also open-world. | ## Workflows, apps, and skills [#workflows-apps-and-skills] | Tool | Title | Reads only | Does | | ------------------------------- | ---------------------------------- | ---------- | ------------------------------------------------------------------------------------------- | | `search_resources` | Search Workflows, Apps, and Skills | yes | | | `load_resource` | Open a Workflow, App, or Skill | yes | Load its files into the sandbox for editing. | | `create_resource` | Create a Workflow, App, or Skill | | Create and scaffold. Runs immediately; see the caveat below. | | `update_resource` | Publish Changes | | Validate and publish the sandbox files as the live version. | | `validate_resource` | Validate Changes | yes | Spec, wiring, and type checks, no publish. | | `get_live_files` | View Published Files | yes | | | `get_resource_config` | View Settings | yes | Credential, const, and trigger bindings. | | `update_resource_config` | Update Settings | | Wire credentials, consts, schedules. | | `set_workflow_type` | Set Workflow Type | | Switch code and agent. | | `set_resource_project` | Move Into a Project | | | | `search_trigger_providers` | Search Triggers | yes | | | `fetch_workflow_trigger_sample` | Fetch Sample Trigger Data | yes | Live records from a configured poll trigger. | | `stage_workflow_input_file` | Attach a Workflow Input File | | Stage a file for a file-type input. | | `preview_workflow_filter` | Preview Workflow Filter | | Run the workflow's `filter.ts` once and return its decision. The only way to test a filter. | | `run_workflow` | Run a Workflow | | Start a test execution in the background. | | `get_workflow_executions` | View Workflow Runs | yes | | | `get_workflow_execution_traces` | View Workflow Run Logs | yes | Input, run record, logs, and the step thread. | | `load_skill` | Load a Skill | yes | Load a workspace or platform skill as context. | ## Apps [#apps] | Tool | Title | Reads only | Does | | ------------------------------ | ---------------------- | ---------- | ---------------------------------------- | | `get_app_build_status` | Check App Build Status | yes | | | `run_app_handler` | Run an App Action | | Run a handler against the live app. | | `run_app_agent` | Run an App Agent | | Start a declared app agent. | | `get_app_handler_invocations` | View App Activity | yes | | | `get_app_errors` | View App Errors | yes | Browser-side errors the iframe reported. | | `invalidate_app_handler_cache` | Refresh App Cache | | | ## Projects and storage [#projects-and-storage] | Tool | Title | Reads only | Does | | ------------------- | ------------------- | ---------- | --------------------------------------------------- | | `list_projects` | List Projects | yes | | | `create_project` | Create a Project | | | | `update_project` | Update a Project | | | | `delete_project` | Delete a Project | | | | `read_project_file` | Read a Project File | yes | A shared file from a project the member can access. | | `storage_list` | Storage List | | Objects in a user store. | | `storage_get` | Storage Get | | Object to a sandbox path. | | `storage_put` | Storage Put | | Sandbox path to an object. | | `storage_delete` | Storage Delete | | | Unlike a workflow run, which reaches only the stores it declares, the connector's storage tools are unscoped: they see every user store the member can reach. None of the four is annotated read-only, `storage_list` included, so a host that gates on annotations prompts for all of them. ## Browser [#browser] | Tool | Title | Reads only | Does | | ----------------------- | ------------------------- | ---------- | --------------------------------------------------------------- | | `browser_session_open` | Open a Browser Session | | A hosted browser the agent drives with Playwright over CDP. | | `browser_session_close` | Close the Browser Session | | Save the login, stop billing. | | `browser_report_login` | Confirm Browser Login | | Mark a saved login `active` or `logged_out`. | | `browser_request_login` | Request a Browser Login | | The live sign-in handoff. Not available over the connector yet. | ## Workspace and memory [#workspace-and-memory] | Tool | Title | Reads only | Does | | --------------------------- | ---------------------- | ---------- | -------------------------------------------------------------------- | | `list_organization_members` | List Workspace Members | yes | Names and emails, for resolving a person to an address. | | `user_question` | Ask a Question | | Comes back as text for you to answer in the chat. | | `remember` | Remember | | Queue a change to workspace or personal memory for the memory agent. | ## Two caveats [#two-caveats] **Nothing on this surface asks you to confirm.** In the dashboard, creating a resource shows a widget with a Create button you press. Over the connector there is no such step: `create_resource` runs the moment the model calls it, and the only confirmation is whatever the assistant chose to ask you in the chat. Worth knowing before setting every tool to always-allow. **The browser tools appear only where the deployment has browser automation configured.** Without it those four are absent rather than failing.