Private sandbox
for
multi-model deliberation
Design, execute and repeat complex deliberation sessions in a secure, browser‑native environment with total data sovereignty.
No registration required.
Revenue Diagnostic
Identify growth gaps and get a structured proposal.
Research Panel
Compare model configurations on the same task.
Expertise-as-a-Service
One engine. Many applications.
Design a session once. Run it with different inputs, different models, different scenarios.
Zero-knowledge by design
Your methodology never leaves your browser.
Workflows execute in a sandboxed runtime. API calls go directly to model providers. The server never sees your data.
Common questions
What models can I use? + −
Any model with an API — Anthropic, OpenAI, Google, AWS Bedrock, or local models via Ollama and LM Studio. You configure the providers in your session.
Do I need API keys? + −
You can bring your own keys or use a Parley subscription that includes model access.
Where does my data go? + −
Nowhere. The runtime executes locally in your browser. API calls are made directly from your browser to your model providers, and we cannot see what you are building or running.
Can I design my own topologies? + −
Yes. Describe the topology in natural language, use the visual editor, or write it in AIR directly.
What is AIR? + −
AIR (Agentic Intermediary Representation) is the specialized language used to define the topologies that run in your browser. You don't need to know it to use Parley, but it's there if you want to edit and run custom topologies directly.
Is Parley open source? + −
The AIR language, compiler and runtime are open source. The Parley Workbench is a hosted product.