Cue Cloud
Cue Cloud

Coding inference, per seat.

Open coding models on hardware we own, tuned for concurrent agents. $1,500/mo per seat · unlimited tokens for that user. Run them in CueCode (our IDE and harness) or hit the same API from whatever IDE you already use.

See how it works →

Product

What Cue Cloud is.

Coding inference we own. Flat seats, unlimited tokens. Open a topic for the full breakdown.

Product topics
cuecloud / product
api.mdOpen full page →

OpenAI-compatible API

OpenAI-compatible. Works anywhere — optimized for IDE coding agents.

  • POST /v1/chat/completions
  • model: "deepseek-v4-pro" | "kimi-k2.7-code" | "glm-5.2"
  • $1,500/seat/mo · unlimited tokens
Read api.md

Why Cue Cloud

Fixed seat · owned decode · unlimited agent tokens

The invoice became the bottleneck.

Frontier APIs price coding agents like chat: every read, tool call, and retry lands on an OpenAI or Anthropic meter. Teams want agents on every desk; finance wants a ceiling. Cue Cloud is that ceiling — because inference runs on Mac Studios we operate, not a thin client over someone else's /v1/chat.

simulate agent loop

Metered frontier

OpenAI / Anthropic · $/MTok

waiting to start…

invoice$0

then metered costs · every turn adds $

Cue Cloud

Mac Studios · flat seat

waiting to start…

seat$0

upfront once · every turn +$0

Illustrative turn costs for one heavy agent session. Not a live invoice. Upfront cost first, then the same loop on both sides. Full monthly math on seats.

Owned hardware is the pricing model

Mac Studio–class nodes. Open Hub weights. We sell seats on capacity we control — not a markup on someone else's rate card. That's how flat unlimited stays real while agents stay concurrent and long-running.

See seats math →

Paths

Two doors. One inference plane.

Owned decode either way. CueCode rides the density path; the API drops into the IDE you already use.

Explore CueCode →
See API →

Same Cue Cloud. CueCode unlocks denser agent signals; the API works everywhere.

Workflow

From access to agents.

Three beats to a flat seat. How the system runs →

  2. CueCode →See API →
  3. See seats math →

Proof

See it decode.

Shaped demo probe — pacing only until the fleet is public.

DEMO · NOT PRODUCTIONdeepseek-v4-prosess_xvuvpj
—
TTFT—
tok/s—
tokens out0

shaped band 16–25 tok/s

warming session…

Demo pacing only. Not a throughput promise or SLA. Band shown is illustrative (16–25 tok/s). Real measured numbers when Cue Cloud is live.

Teams

Flat seats. Bills you can explain.

$1,500/mo per seat · unlimited tokens for that user. Team keys live in the Cue Cloud console. Ships with access.

Ready to run on Cue Cloud?

change theme