Media tools
Search · vision · image · audioProvider-backed sandboxes. Try them here, then hand to agents.
What this is
The four media-shaped tools: real web search, vision analysis, image generation, TTS.
What to do
Pick one. Pages tagged NEEDS PROVIDER need a key in .env.local before they'll do real work.
What happens next
Calls run through the same registry agents use, so anything you test here works the same inside a mission.
Web search
READYReal search via Tavily / Brave / SerpAPI. Returns ranked results with title, url, snippet, source.
Vision analyze
READYSend an image URL + a task to a multimodal Hermes model. Returns description, OCR text, findings.
Image generation
READYGenerate an image via an OpenAI-compatible /images/generations endpoint. Approval-gated when used by agents.
Text-to-speech
READYSynthesize audio via an OpenAI-compatible /audio/speech endpoint. Returns base64 audio.


