Queue and pace requests
Absorb bursts and release work as configured request, token, spend, and shared-capacity windows open.
Run the queue-first scheduler, limits, routing lanes, fallbacks, and embedded dashboard on your own infrastructure.
POST /v1/chat/completions
lane local-first
queue 2 waiting · 1 active
route local-endpoint
status completed
Put queues, limits, route order, fallback behavior, and request visibility between your applications and compatible model endpoints.
Absorb bursts and release work as configured request, token, spend, and shared-capacity windows open.
Set global, provider, endpoint, and lane limits for requests, tokens, spend, and concurrency.
Group compatible endpoints into ordered paths and keep local, cloud, or aggregator routes behind one API.
Track cooldowns, requeue throttled work, and hand supported failures to the next configured path.
Estimate usage before dispatch, record provider-reported tokens, and apply your configured endpoint pricing.
Configure paths, inspect the queue, review requests, and understand usage from the embedded admin UI.
Run the Go service, connect compatible endpoints, order them into lanes, set limits and pricing, then point your applications at AnchorShell Relay.
AnchorShell Relay encrypts provider credentials with a local master key. Operational metadata gives you a useful request trail, while request and response body retention remains an explicit operator setting that defaults off.
Use the verified AnchorShell reporting path for security issues. Keep provider keys, access tokens, and sensitive model payloads out of an initial report.
Security reportingAnchorShell Relay — Managed adds organization-aware access, permissions, per-user guardrails, and managed operation to the same traffic-control architecture.
| Area | Relay — Self-Hosted | Relay — Managed |
|---|---|---|
| Operation | You run the single-node service | AnchorShell runs the managed product |
| Configuration data | SQLite by default | PostgreSQL-backed organization data |
| Identity | Local administration | Platform users, organizations, and permissions |
| Guardrails | Shared Relay resource limits | Shared limits plus user/model/provider policies |
| Traffic-control core | Queue, pace, route, fallback, observe | The same core with managed controls |
Inspect the traffic-control runtime, follow the self-hosting docs, and use the repository’s contribution and issue paths.
Self-host AnchorShell Relay or get started with the organization-aware managed service.