# OpenRTC

## Docs

- [OpenRTC](https://docs.openrtc.tech/index.md): livekit-agents on steroids. Host all your agents in one worker, with hot reload, live introspection, per-tenant isolation, and zero-downtime deploys. Your Agent code does not change.
- [Coming from livekit-agents](https://docs.openrtc.tech/coming-from-livekit-agents.md): You already write livekit-agents. OpenRTC hosts all your agents in one worker and adds hot reload, per-tenant isolation, and zero-downtime deploys, without changing how you write an Agent.
- [Getting Started](https://docs.openrtc.tech/getting-started.md): Install OpenRTC and run your first multi-agent LiveKit worker.
- [Examples](https://docs.openrtc.tech/examples.md): Multi-agent example showing how to register, route, and run two agents in one OpenRTC worker.
- [Architecture](https://docs.openrtc.tech/concepts/architecture.md): How OpenRTC's coroutine-mode worker, session lifecycle, and shared prewarm work.
- [Frameworks](https://docs.openrtc.tech/concepts/frameworks.md): Run OpenRTC on livekit or pipecat through one backend seam, and how agents register on each.
- [Routing](https://docs.openrtc.tech/concepts/routing.md): How OpenRTC resolves which Agent handles each incoming session, with priority rules, metadata formats, and error behavior.
- [Hot Reload](https://docs.openrtc.tech/concepts/hot-reload.md): Edit an agent file and OpenRTC swaps live sessions to the new class on their next turn, with no dropped calls.
- [Session Introspection](https://docs.openrtc.tech/concepts/session-introspection.md): How OpenRTC attributes memory, CPU, and event-loop blocks to individual sessions inside one shared worker, what it measures, and what it cannot see.
- [Multi-tenancy](https://docs.openrtc.tech/concepts/multi-tenancy.md): Run many clients (tenants) in one OpenRTC pool with per-tenant provider keys, resource fairness, and blast-radius isolation, plus what is shared and what is not.
- [Migration and drain](https://docs.openrtc.tech/concepts/migration.md): Why OpenRTC upgrades workers by draining live calls rather than migrating them, what state can and cannot move, and when migration might return.
- [Zero-downtime Deployments](https://docs.openrtc.tech/operations/deployments.md): Upgrade an OpenRTC worker fleet without dropping calls, using blue-green drain.
- [Monitoring a deploy](https://docs.openrtc.tech/operations/monitoring-deploys.md): The OpenRTC runtime signals to watch while a blue-green deploy is in flight, and where the cost and quality signals live instead.
- [Rolling back a deploy](https://docs.openrtc.tech/operations/rollback.md): Undo a bad OpenRTC version the same way you rolled it out, by draining, plus a decision tree for when to roll back versus fix forward.
- [Audit events](https://docs.openrtc.tech/compliance/audit-events.md): The OpenRTC deployment audit-event reference, with the schema, event types, sample SIEM queries, and how to provision the signed-membership secret.
- [Debugging Density](https://docs.openrtc.tech/runbooks/debugging-density.md): My pool feels slow, or one session is hot. A troubleshooting flow using openrtc top and the slow-session detector.
- [Onboarding a Tenant](https://docs.openrtc.tech/runbooks/onboarding-a-tenant.md): The step-by-step flow to add a new client (tenant) to an OpenRTC pool safely.
- [Tenant Incident](https://docs.openrtc.tech/runbooks/tenant-incident.md): Identify, isolate, and escalate a misbehaving tenant without disrupting the others.
- [CLI Reference](https://docs.openrtc.tech/cli.md): All openrtc subcommands, flags, runtime knobs, and metrics output options.
- [openrtc top](https://docs.openrtc.tech/cli/top.md): htop-style live inspector for the shared-worker session pool (columns, key bindings, filters, and flags).
- [AgentPool](https://docs.openrtc.tech/api/pool.md): Full API reference for AgentPool, AgentConfig, add(), discover(), run(), and routing.
- [Density Benchmark (v0.1)](https://docs.openrtc.tech/benchmarks/density-v0.1.md): 50+ concurrent sessions per worker at under 4 GB peak RSS (methodology and results).