Questions about APEXVYRA

A control plane for AI agents and workflows: you build an agent once, and APEXVYRA handles routing it to a model, running its tools, and logging what happened.

Most teams send their first request within an hour of signing up — create a workspace, generate an API key, and POST to /v1/runs. No orchestration code required for a single-step agent.

You define what an agent can do (its tools and where it pulls context from) and, optionally, a routing policy. If you skip the routing policy, APEXVYRA uses a sensible cost/latency default.

Model usage. Each run consumes credits proportional to tokens processed and the model tier it was routed to — a Fast-tier run costs fewer credits than a Frontier-tier run of the same length.

Yes. Upgrading or downgrading changes your credit pool, seats, and feature access immediately; nothing about your existing agents, workflows, or run history is affected.

Requests beyond your plan's requests-per-second limit (5 on Free, 20 on Starter, 100 on Scale, custom on Enterprise) queue briefly and retry automatically rather than failing outright.

7 days on Free, 30 on Starter, 90 on Scale, and configurable up to 7 years on Enterprise. Retention is set per workspace.

SSO/SAML is included on Scale and Enterprise; Enterprise also includes SCIM for automated user provisioning.

Access follows your workspace's role-based permissions — only members with run-visibility access can view prompts, outputs, and traces, and every view is itself part of the audit log.

First-party SDKs for Python, Node, and Go, plus a plain REST API for everything else.

Yes — see the Integrations page for the current connector categories, including databases, data warehouses, and generic webhooks for anything without a dedicated connector.

Every workspace includes an evaluation suite for scoring a prompt version or agent against saved test cases before you promote it to production traffic.