Tiers¶
Hive's tiers are distinguished by which relay you connect to, not by which app you run. The client is the same OSS binary everywhere; moving between tiers is a connection change in Settings → Multiuser sync (and, for paid tiers, an entitlement on your org).
| Tier | Who runs the relay | Cost | Notes |
|---|---|---|---|
| Self-host | you, on your hardware | free (MIT) | full core function |
| Hive Cloud — Free | Hive-operated | free | 1 team, ≤5 members, 30-day retention, best-effort forwarding |
| Hive Cloud — Pro | Hive-operated | ~$6/mo | more members, 1-year retention, guaranteed TURN |
| Hive Cloud — Team | Hive-operated | ~$4/user/mo | unlimited members, admin controls + audit, org/seat management |
| Enterprise | you, on-prem | site license | SSO, audit export, DLP hooks, data residency, support |
Self-host (free, MIT)¶
Run hive-relay yourself — the full source is in this repo. See
Self-hosting a relay and the
small-team deployment guide.
Hive Cloud¶
A managed, multi-tenant relay. Same protocol as self-host — the client doesn't change; you just connect to a Hive-operated relay (and, for paid plans, paste a relay access token in Settings → Team sync).
What the relay actually does (why it's cheap)¶
The relay is sync + broker only. It forwards end-to-end-encrypted event envelopes, brokers short pairing codes, holds sealed key-rotations, and resolves GitHub-handle → device lookups. LLM inference never goes through the relay — the client talks to your model provider directly. So Hive Cloud has no token cost; pricing reflects convenience and scale, not compute.
The only resources that scale with use are relay-forwarded bandwidth when peers can't connect directly (TURN fallback) and history retention (how far back the relay can replay events to a brand-new device). Plans meter those.
Plans¶
| Free | Pro | Team | |
|---|---|---|---|
| Price | $0 | ~$6/mo flat | ~$4/user/mo |
| Hosted team workspaces | 1 | a few | unlimited |
| Members per team | up to 5 | up to ~15 | unlimited |
| Devices per account | unlimited | unlimited | unlimited |
| Backfill retention | 30 days | 1 year | unlimited |
Invite by GitHub @handle |
✓ | ✓ | ✓ |
| Direct P2P (always free) | ✓ | ✓ | ✓ |
| Guaranteed TURN forwarding | best-effort | ✓ | ✓ |
| Team admin controls + audit | — | — | ✓ |
| Org/seat management, policy templates | — | — | ✓ |
Solo / local workspaces are always free and unlimited — they never touch a relay. Retention limits only affect how far back the relay can replay to a new device; every existing device keeps its full local history forever.
We don't meter devices. Using the same account on your laptop and your desktop is core to how Hive identity works, not an upsell.
Numbers above are indicative, not a committed price sheet. The Team tier's admin
controls (fine-grained roles, server-enforced membership, audit trail) are the
main paid surface and are still being built — see
docs/managed-service-plan.md in the repo.
Enterprise¶
An on-prem relay with full role-based access control and a support contract. The value-add sits next to the same protocol — your team runs the same OSS client.
- RBAC — capability-based roles (not just owner/admin/contributor/viewer): define custom roles from grants like invite, remove member, rotate key, approve execution, manage integrations, view audit.
- GitHub Enterprise Cloud (GHEC) — sign in through your GHEC org's SSO (SAML/OIDC via Okta / Entra / Google); map GitHub teams → Hive roles, so provisioning and deprovisioning follow GitHub. GitHub Enterprise Server (self-hosted) is supported too.
- Governance for agent actions — policy over what agents may do to a shared tracker (file issues but not close them), gated execution, plus an audit trail of every membership, key-rotation, and agent-made change, exportable to S3/SIEM.
- On-prem + offline licensing — the relay verifies a Hive-signed license with a public key, with no callback to Hive. Data-residency follows wherever you run it. DLP hooks available.
What stays free forever¶
- The Hive client (all OSes).
- The reference relay (MIT) — envelope forwarding + end-to-end encryption.
- The protocol specifications.
A team that never touches Hive Cloud can run its own relay forever.
For the engineering model — how each feature gates (server capability vs runtime
entitlement vs build flag) and where it lives — see docs/tiering.md in the
repository.