983 / 983 tests pass

One of four products in the A.L.I.C.E. ecosystem

Runtime Single-binary AI agent gateway

Multi-model. Multi-channel. Multi-tenant. Open core.

An opinionated, single-binary AI agent gateway with native memory, multi-channel delivery, three-tier inference, and a real intelligence loop. Drop it onto a Mac or Linux box, point any frontend at it.

What's built

An opinionated, single-binary AI agent gateway.

  • OpenAI- and Anthropic-compatible HTTP APIs
  • MCP server + client
  • Three-tier inference: direct adapters, LiteLLM sidecar, llama.cpp local with CoreML/ANE
  • 7 native channel adapters (Telegram, WhatsApp, Slack, Discord, Google Chat, Teams, WebChat)
  • Multi-agent + observable, cancellable sub-agents
  • Governed learning loop with eval gate and rollback
  • Merkle-chained tamper-evident audit log
  • Cloud-sandbox execution (Daytona, Modal, Vercel, K8s)

Runtime pricing

Open core. Five tiers.

Buy Runtime alone, or save 30% with the A.L.I.C.E. | Stack bundle. See full pricing →

Open Core

$0MIT

Single binary, all APIs, basic learning loop

Team

$99/mo

Priority support, multi-tenant RBAC, audit log

Business

$499/mo

SSO, advanced RBAC, governed skills, scheduled ops

Enterprise

$2,500+/mo

On-prem guides, FedRAMP-prep posture, dedicated CSM

Sovereign

$10,000+/mo

Air-gap, FedRAMP/IL5, classified-environment hardening

The A.L.I.C.E. ecosystem

Four products. One stack. A-la-carte or all four.

Each product works independently. Together they compound — and the bundle is 30% off list.