One list. Priced per governed agent.

Is cond_agt_* priced per user or per bot?

Per bot — per governed service identity. One human developer running three governed agents counts as three identities. The unit tracks value produced by the agents, not headcount using the tool.

Does self-hosted include air-gapped deployments?

Docker self-hosted ships today; Kubernetes reference templates are in preview; fully air-gapped installs are running with design partners and not yet generally available. Same image, same policy engine, same list price as SaaS — you’re paying for enforcement and audit, not where the pods run. Talk to us if you need the air-gapped track before it opens up.

What if my agent count varies month to month?

Billed on the peak count of active identities in each month. Spin agents up for a launch, retire them after, and you only pay for the month you actually ran them. No proration penalty for scaling down.

Do you offer annual discounts?

Yes — 15% off list price on all tiers when paid annually up front. Available on Team, Business, and Enterprise. Ask for the annual quote when you talk to us.

What happens if I hit my tier cap?

Soft alert at 90% of cap so you can plan the upgrade. If you hit 100%, agents keep working — we auto-upgrade you to the next tier and send a confirmation email. No hard stops, no surprise blocks, no missed enforcement.

How is our data stored and processed when it passes through third-party LLMs?

Every call flows through the Conduct proxy first: workspace-scoped vault keys (no shared credentials), pre-flight redaction of PII, secrets, and path prefixes, and a policy check that decides allow / block / approve / audit before the request leaves your perimeter. Every decision writes a SHA-256 hash-chained audit receipt. Deployment options: SaaS shipped, Docker self-hosted shipped, Kubernetes preview, air-gapped design partner program.

What happens to our margins when model prices jump 2–3× overnight?

Zero direct exposure. Customers bring their own model keys — Conduct never resells tokens. We charge per governed decision, not per token, so model price swings do not move your unit economics. Cross-provider routing is native: if one vendor spikes, workspace policy can route to a cheaper equivalent without any application change.

What happens if Claude Code, OpenAI, or Microsoft ships a native version of this?

Vendor-native governance is single-vendor by definition — Anthropic will not ship enforcement for OpenAI models, OpenAI will not ship it for Anthropic, and neither will govern the MCP tools your agents actually call. Conduct is the cross-cutting layer: one policy across every model, agent, and tool your enterprise runs. That heterogeneity is structural, not a feature race.

Why leave the major platforms for you?

One policy across N vendors, M agents, and K tools — a scope no single platform can offer without cannibalizing itself. Hash-chained audit that legal and SOC 2 auditors accept as evidence. Apache-2.0 core with patent-pending enforcement (US 64/109,502), so if we disappear tomorrow the runtime keeps running. Vendor logs are marketing; hash-chained receipts are proof.