IO is not a single-model chat wrapper. It is a private AI provider — an inference layer reachable from web, CLI, MCP, SDKs, and messaging relays. Every surface inherits the same three properties: open-weight multi-model inference, redaction before the model observes input, and a signed receipt on every response.
Multi-Model Private AI
No account, no sign-in. Solo, Broadcast, Diffuse, or Fork — with self-destruct timers on every session.
Multi-Path Coding Agent
Proposes several solutions in parallel. Fork, explore, merge. Redact secrets before any model sees them.
Orchestration Layer
21 tools spanning core inference, multi-model orchestration, privacy & proof, session control, collaboration, and audit.
Embed the provider into your own stack. Rust, Python, TypeScript, Go — five lines to inference.
from io_sdk import IO
io = IO()
r = io.completion("Explain ZK proofs",
model="claude",
ttl=300, redact=True)
print(r.text, r.receipt.verify())Extending the same primitives to new surfaces.
E2E encrypted collaborative rooms with private AI sidebar
Infinite whiteboard with AI co-creator
Document analysis with field-level redaction
Client-side encrypted agent memory
Autonomous agents with receipt-audited steps
Telegram, WhatsApp, Signal, Discord, SMS bridges