Skip to main content
Provider tools are first-class capabilities that ship with HasteKit. Toggle them on in the Tools tab and the agent gets a new tool — no MCP server, no OpenAPI spec, no custom code. Each tool has independent Enabled, Requires Approval, and Deferred switches.
Tools tab

​
Available tools

ToolWhat it doesConfigurable
Image GenerationGenerate images from text prompts.Provider + model (e.g. xAI grok-imagine, OpenAI gpt-image-1).
Speech Generation (TTS)Synthesize speech from text.Provider + model + voice + language (e.g. ElevenLabs Flash v2.5).
Transcript Generation (STT)Transcribe audio to text.Provider + model (e.g. ElevenLabs scribe_v2).
Web SearchSearch the web for real-time information.None — uses the provider’s native search.
Code ExecutionRun code in the provider’s sandboxed code interpreter.None.
Todo ToolA built-in todo list the agent uses to plan and track multi-step tasks.None.
Schedule Task ToolLets the agent schedule itself for later — recurring (cron) or one-off — and amend existing schedules.None.
Sandbox ToolPer-session Docker-backed bash sandbox with persistent workspace and file API.Image, env vars. See Sandbox Tool.

​
Enabling tools

  1. Open your agent in the Agent Builder.
  2. Click the Tools tab.
  3. Toggle the tool on. Image / Speech / Transcript tools also let you pick the provider and model.
  4. Optionally flip the Requires Approval switch to gate each tool call behind human-in-the-loop confirmation.
  5. Optionally flip Deferred to keep the tool definition out of the prompt until the agent explicitly asks for it (useful for big tool catalogs).
  6. Save.

​
Image / Speech / Transcription

These three are configured similarly: pick a provider, pick a model, and (for speech) a voice and language. The agent receives a single tool — e.g. generate_image(prompt), generate_speech(text), transcribe(audio_url) — and the runtime routes the call to the configured provider behind the scenes. You can use any combination of providers for the three tools — e.g. xAI for images and ElevenLabs for speech.

​
Web Search & Code Execution

These map onto the LLM provider’s native capabilities. Compatibility depends on the model you pick on the Model tab.
ToolOpenAIAnthropicGemini
Web Search✅✅—
Code Execution✅✅✅
For richer or stricter sandboxing — or to give the agent a persistent workspace — use the Sandbox Tool instead of provider-level code execution.

​
Todo & Schedule

Two utility tools designed for long-running, agentic workflows:
  • Todo Tool — gives the agent add_todo, update_todo, list_todos. Useful for keeping a complex plan organized when the agent is doing many things in a single run.
  • Schedule Task Tool — gives the agent schedule_once(at, message) and schedule_recurring(cron, message) so an agent can defer work, set reminders, or kick off recurring runs of itself.
Both compose naturally with Triggers — the agent can register a trigger and later be invoked by it.