AiMedley is where personas outlive vendors. Create persistent AI workers — give each a name, mission and personality — then pick which of 12 brains across 8 providers thinks for them today. Swap models tomorrow without losing anything.
Same prompt. 2–5 workers. Each argues a different stance. A judge worker weighs the arguments and delivers a structured verdict — live, in parallel, in under a minute.
Every LLM release resets your setup elsewhere. Not here. Your worker keeps their name, mission, memory, and history no matter which brain you swap in.
12 models across 8 providers behind one unified interface. Bring your own OpenAI / Anthropic / Google keys to bypass the middleman and cut cost 30-70%.
Give the same prompt to 5 workers with different stances. A judge weighs the arguments in a live streamed verdict. Every debate is shareable in one click.
Workers, conversations, debates, and API keys are scoped to your account. Direct provider keys are encrypted at rest. No training on your prompts. No ads.
Every brain behaves the same way to your workers — swap them like keyboard shortcuts. Bring your own keys where you can to cut cost and latency.
Create AI colleagues with names, missions, memories, and expertise. Chat with each independently.
Change which LLM powers a worker anytime. Persona, memory, and history are preserved.
2–5 workers argue their stances. Judge worker delivers a structured verdict. Live SSE streaming.
Group workers under a leader with shared instructions and memory. Broadcast prompts to the whole team.
Save any worker as a template. Clone the community's best setups in one click. Publish yours.
Templates, conversations, and debates become no-auth URLs. Snapshot-based — no live LLM leak.
Fan the same prompt out to up to 4 brains simultaneously and watch them stream side-by-side.
Store your OpenAI/Anthropic/Google keys once (encrypted). Every worker on those brains bypasses OpenRouter.
Daily and monthly caps per worker or globally. Live spend counter. Warn at 80%, block at 100%.
The moment you sign up we spin up 4 demo workers + 1 team so you can run a debate immediately. No setup, no config, no prompt engineering.
Behind every worker is a unified `chat_stream(brain_id, msgs)` interface. Switch a worker from GPT-5.1 to Claude Sonnet 4.6 to Gemini 2.5 Pro with a dropdown.
Save your OpenAI / Anthropic / Google keys once. Workers on those brains bypass OpenRouter and use the vendor natively — cheaper, faster, no middleman margin.
Each round fans debaters out in parallel via server-side asyncio queues. Once every debater finishes, the next round starts. The judge sees the full transcript and streams a structured verdict.
When you share a template, chat, or debate we freeze the payload. Public viewers never re-trigger LLM calls or leak live worker state. You can safely edit the source after.
Spin up a 4-worker team in 30 seconds. Run your first debate for a penny. No card, no lock-in.
Create my workforce