Skip to content

Hivemind

Hivemind is in early alpha

Hivemind is currently being deployed to design partners only as we shake out the install + ops story on real customer infrastructure. Public availability and pricing follow once the install + upgrade flows are fully self-service.

Want to be a design partner? Email sales@seglamater.com.

What is Hivemind?

Hivemind is a multi-agent AI orchestration server you run on your own hardware. It coordinates a fleet of specialized AI agents — research, code, ops, custom — against the model providers you choose (Anthropic, OpenAI, Ollama on your own GPU, or any OpenAI-compatible endpoint), and exposes them through a single conductor agent that can plan multi-step work and route subtasks to the right specialist.

Grounded support bot

The first thing most teams build on Hivemind is a grounded support bot for their Chatalot community. You create an agent, load a knowledge base of your facts, and connect it to a community — it then answers questions in DMs and (optionally) when @-mentioned in channels, grounded in your knowledge so it answers from your facts and says "I don't know" instead of inventing. The bot runs entirely on your infrastructure, on the model you choose, behind your SSO. See the Chatalot integration guide for the end-to-end setup.

Why self-host?

  • Your data, your hardware. Conversations and agent context live on the server you control. No telemetry, no SaaS round-trip.
  • Bring-your-own models. Pay the API providers you already pay, use the local models you already host, mix and match per agent.
  • Bring-your-own SSO. Hivemind sits behind whatever identity provider you already run.
  • A real ops story. Signed releases, pinned image digests, reproducible installs. Not a Jupyter notebook in a tmux pane.

How is it shipped?

A single signed bundle that brings up a small Docker Compose stack (Postgres + the hivemind-server API + a small web UI). One install command, one config file at /srv/hivemind/.env, idempotent re-runs.

Channels

Channel Audience Cadence
canary Design partners Per release
beta Not yet populated —
stable Not yet populated —

Existing customers — see the Quick start for the install one-liner and Operator notes for upgrades, backups, and troubleshooting.

Detailed docs

  • Architecture — what each component is and where it sits on your host.
  • Install — fresh-install walkthrough.
  • BYOK LLM setup — provider snippets for Anthropic / OpenAI / Ollama / OpenAI-compat.
  • SSO setup — Authentik / generic OIDC.
  • Upgrading — managed-update path + manual fallback + rollback.
  • Troubleshooting — boot-time and runtime failure modes + fixes.
  • Bundle spec — the per-customer signed bundle.json that drives every install.
  • API reference — the customer API surface (auth, chat, integrations, updates).
  • Chatalot integration — stand up a grounded support bot end-to-end (provision, knowledge base, channel @-mentions) plus the outbound agent-tool surface and TLS posture.
  • License — proprietary terms, no-telemetry posture, contacts.