Run agents that show up for you.
Multi-agent runtime with shared memory, per-tenant tools, full tracing. Discord, REST, and CLI on top of one process.
Real screens from a seeded Catentio workspace — not a marketing render. Click through the tour below to see what the portal looks like once data is flowing.
Step 01
A multi-agent runtime: one process, many personas. The classifier reads incoming traffic and routes to the right agent, then the runtime executes against per-tenant tools and emits a tracing event for every call. Built for indie operators running their own AI ops, not a labs project.
Public preview — registration opens later in 2026
Step 02
Per-agent system prompts, BYO model keys (Anthropic, OpenAI, Groq), scoped tool surface, output destinations (Discord, REST, CLI), and full request/response tracing with replayable Run IDs.
Step 03
Usage-based on top of your own model keys. The runtime charges per agent-hour active; you pay model providers directly. No platform tax on tokens.
Step 04
Architecture overview, agent definition syntax, tool registration, deployment guide, and the REST + Discord surface reference. Concrete and code-first.
One process, many personas. Classifier routes; runtime executes.
BYO tokens, scoped tool surface, output destinations.
Tool + model calls emit events; Run IDs propagate; logs replayable.