Services beyond Beelab
Custom AI on your docs. Migrations from any cloud. Hands-off support. Enterprise installs.
Your AI, on your hardware
Three scope layers covering the AI gateway, RAG and CAG patterns, the memory layer, agent automation, and the on-prem inference layer. Beelab Studio carries an AI baseline (gateway + Open WebUI + on-prem inference); Beelab Pro adds MCP + eval + memory; Beelab Enterprise adds custom MCP + per-team keys + on-prem inference at scale. There is no separate AI SKU. Buyer brings their own model keys. The infra is yours. Final scope is confirmed in a signed SOW before any payment; out-of-scope work is quoted at a documented hourly rate.
AI gateway supply-chain note: The AI gateway layer is pinned to verified post-incident versions and runs inside your firewall. The March 2026 PyPI supply-chain incident affected gateway versions v1.82.7 and v1.82.8; we pin to a patched release (v1.83.0+) and verify install hashes before shipping. source.
The AI baseline ships inside Beelab Studio; advanced AI ships inside Beelab Pro and Enterprise. Email me before ordering so we can confirm scope and timing.
AI Baseline
Included in Beelab Studio
The AI baseline: OpenAI-compatible gateway, Open WebUI front-end with SSO, and on-prem inference. Buyer brings their own model keys or supplies Apple Silicon hardware for local inference.
- OpenAI-compatible AI gateway (default routing layer, yours to swap)
- Chat front-end with single sign-on (Open WebUI + Authentik by default, swap front-end or IdP on request)
- On-prem inference layer (Ollama default, supplied by you if no Apple Silicon hardware; swap for vLLM or llama.cpp on request)
- Buyer supplies API keys for any one of: Bedrock, OpenAI, Anthropic, Vertex, Azure, or local Ollama (current vendor list, subject to upstream API stability)
- RAG-ready data path documented for the buyer's data
AI Workspace
Included in Beelab Pro
Adds MCP connectors from a published supported list, an eval harness, and the self-hosted memory layer. Scoped automations, audit, and traces agreed during scoping.
- Everything in AI Baseline
- MCP connectors from a published supported list (vendor or community, subject to upstream availability)
- Self-hosted memory layer (mem0 or equivalent, current default subject to upstream availability)
- Activepieces with buyer-defined automations, each scoped to documented connectors and a build budget agreed in the SOW
- Audit logging plus traces via Langfuse or equivalent (current default subject to upstream availability)
- Eval suite with pass criteria defined and agreed during scoping
- RAG and CAG patterns documented for the buyer's data
AI Sovereign
Included in Beelab Enterprise
Adds a custom MCP server, virtual per-team API keys, on-prem inference scaled across multi-node Apple Silicon mesh, and a 90-day post-deploy retainer with capped hours.
- Everything in AI Workspace
- One custom MCP server for one buyer system, scoped to a documented hour ceiling in the SOW
- Virtual API keys per team with scoped budgets and quotas
- On-prem inference scaled across multi-node Apple Silicon mesh (Ollama or vLLM, default; swap on request) sized against a documented model menu
- Temporal memory layer (current default, swap on request)
- 90-day post-deploy retainer, hours capped at the scoping document
- Hardware purchase remains the buyer's call; sizing recommendation provided against the documented model menu
Industry pricing for comparable RAG and MCP integrations typically falls in the mid-five-figure range; pricing varies by scope. Beelab includes the infra, so you pay for the integration, not the stack. Every tier runs on hardware you already own. No SaaS fees, no token tax, no vendor lock.
Off the cloud, onto your hardware
Most single-tool jobs finish in a day; multi-tool migrations are scoped on a call. Custom and complex migrations are quoted after scoping. Per-migration and bundle pricing is on the hardware page; email me with your source data details to get started.
View migration pricingGoogle Photos
Immich
Full library, albums, dates, EXIF preserved via immich-go.
Dropbox
Nextcloud
Folder tree, sharing perms, version history kept where possible.
Bitwarden Cloud
Vaultwarden
Vault, organisations, attachments, sends. In most cases users do not need to re-authenticate.
Google Drive
Nextcloud
Documents, shared drives, owners remapped to your SSO identities.
Gmail
Stalwart Mail
IMAP sync, labels to folders, MX cutover with old-mail forwarding.
Notion
AppFlowy
Pages, databases, embeds. Markdown export plus structural rebuild.
1Password
Vaultwarden
1pux export converted, vault structure mirrored, MFA seeds carried.
iCloud Photos
Immich
Live Photos, bursts, RAW. iCloud download plus immich-go upload.
Custom migration
Talk to me
Anything not on the list. Scoped on a 20-minute call.
OpenCloud Sidecar
Install OpenCloud, the Heinlein Gruppe Go fork of ownCloud, on an enterprise PC you already own (Mac mini, Beelink, NUC, MINISFORUM). Runs in under 100MB of RAM, talks to your Beelab Authentik, and gives you a quiet second cloud that survives if the rack goes down.
- OpenCloud install on your enterprise PC (Docker or native binary)
- OIDC tied to your existing Beelab Authentik
- 1 desktop client + 1 mobile client setup
- Tailscale install for off-network access
- 60 minutes total on a tested box. Async over email at [email protected], no call required.
Source on the runtime footprint: xda-developers, OpenCloud lightweight Nextcloud alternative.
One-time add-on
Same install, no surprise rebill. Bring your own enterprise PC. Requires an existing Beelab cluster with Authentik configured.
View on pricingOr email me with questionsBigger rack, bigger problem? Talk to me.
For multi-rack or multi-site builds, GDPR or SOC2 readiness review (not legal advice, you will still need your own counsel), or integrations that go past the standard RAG tiers. No fixed price, every engagement is scoped on a call.
- Architecture review
- Custom integrations
- GDPR/SOC2 readiness
[email protected] · best-effort reply within 1 business day
or email me first, I reply within 1 business day.