LoginSign Up
← Back to all sparks
H

Hatz AI

SUPPORT
Velocity6.3

Hatz AI is building the AI workspace for MSPs — per-message model routing, tenant tooling, custom MCP.

mspauto-llm-routingmulti-tenantcustom-mcpworkflows
◆Current state
Hatz AI is shipping at a high cadence across three connected themes. First, model routing: Auto-LLM picks the right model per message based on task and tools, then layered into Lite, Performance, and Turbo tiers; the catalog keeps adding models (Opus 4.7, Gemini 3.5 Flash, Gemini 3.1 Flash Lite, Gemma 4) with per-model credit multipliers surfaced in the UI. Second, MSP control plane: bulk tenant creation via CSV, custom roles with credit limits, workshop access controls, and embedded support chat in the admin dashboard. Third, surface expansion: audio uploads with auto-transcription, image generation in workflows, file output attaching to chats, 60+ supported file types, speech-to-text in chat, and a steady cadence of integrations and custom MCP server improvements.
◆Where it's heading
The product is taking shape as a multi-tenant AI workspace tuned for MSPs and partner-led delivery — the tenant CSV, credit limits, and workshop sharing are unusual for a generalist AI tool and tell you who buys this. Auto-LLM and tiered routing make sense in that context: an MSP needs cost control across many tenants without micromanaging model picks. Custom MCP and the broad integration cadence position Hatz as a tools-aggregator over multiple LLMs rather than a model wrapper.
◆Prediction
Expect more MSP-centric controls — per-tenant budgets, white-label theming, billing reconciliation — and Auto-LLM to grow visible routing telemetry so MSP admins can see why a given model was picked. The custom MCP surface is likely to evolve toward a marketplace pattern with shareable MCP packages across tenants.

◆Recent moves

  1. 1d ago

    Gemini 3.5 Flash Added to Model Selector

    Gemini 3.5 Flash joins the model selector for chat, tool use, and image prompts. Routine catalog addition that also feeds Auto-LLM's routing pool.

    View source ↗
  2. 6d ago

    Continue-in-chat, model multipliers, 60+ file types

    Bundled release: workflows can hand off to a seeded chat, the model selector now shows credit multipliers and promotional models inline, generated files auto-attach to chats and workflows, and supported file types expand to 60+. Everyday-UX heavy, no directional shift.

    View source ↗
  3. 13d ago

    Audio uploads with auto-transcription, 4 new integrations

    Hatz accepts audio uploads (mp3, m4a, wav, etc.), auto-transcribes them, and makes the result searchable across chat, agents, and workflows. Four new integrations (including Pylon for B2B support) and custom MCP improvements ship in the same release.

  4. 20d ago

    Auto Model Selection gets Lite, Performance, and Turbo tiers

    Auto Model Selection now exposes Lite, Performance, and Turbo modes so users pick the speed/cost/capability balance once instead of per-message. The natural tiering layer on top of the Auto-LLM router.

    View source ↗
  5. 27d ago

    Auto-LLM: per-message model routing across providers

    ⚡ SPARK

    Auto-LLM analyzes each message in real time and routes to the most capable and cost-efficient model based on task type, complexity, files, and tools. New entrants Gemma 4 and Gemini 3.1 Flash Lite plus eight integrations land alongside.

    View source ↗
  6. 1mo ago

    Bulk tenant CSV, in-app support chat, workflow cancel, Opus 4.7

    MSP-flavored release: CSV bulk-create tenants, embedded Lark support chat with auto-context, cancel active workflow runs from the run views, seven new integrations, and Opus 4.7 in the model catalog. Operational scale-out work rather than a new direction.

    View source ↗