productapi.md

API

Point any OpenAI-compatible client at Cue Cloud. Your cue_… key works broadly. This stack is optimized for coding inside IDEs and agent loops. Pick DeepSeek V4 Pro 0813, Flash 0731, Kimi K2.7 Code, or GLM 5.2 by model id.

$500 / month = 1 pack = 1B tokens

One cue_… key per workspace. $500 / pack / mo = 1B tokens this month. Agent loops spend include, not a climbing meter. See packs →

Request lab

https://api.cuecloud.io/v1

curl · deepseek-v4-pro

curl https://api.cuecloud.io/v1/chat/completions \
  -H "Authorization: Bearer cue_••••••••" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-pro",
    "messages": [
      { "role": "user", "content": "fix the failing test" }
    ]
  }'

Auth with cue_ keys

Requests carry Authorization: Bearer cue_… from the console. Plaintext shows once when created; the gateway stores a hash. Same key works in an IDE, a script, or CI.

Models by id

Open Hub checkpoints we deploy. Full specs and benchmarks on models.md.

Id	Model	Best for	Source
`deepseek-v4-pro`	DeepSeek V4 Pro 0813	Default agent loops, SWE workloads, long-context coding	Hub →
`deepseek-v4-flash-0731`	DeepSeek V4 Flash 0731	Faster agent loops where Flash quality is enough	Hub →
`kimi-k2.7-code`	Kimi K2.7 Code	Long-horizon coding agents, MCP tool loops, efficient thinking	Hub →
`glm-5.2`	GLM 5.2	Multi-hour agent sessions, planning, 1M-context engineering	Hub →

Point your client

Drop Cue Cloud in as an OpenAI-compatible endpoint. Optimized for coding inside IDEs and agent loops.

Cursor

Add an OpenAI-compatible provider. Base URL → Cue Cloud. API key → your cue_… key. Model → one of the flagship ids.

Claude Code

Point OpenAI-compatible config at the Cue Cloud base URL with your cue_… key. Same wire shape, coding-first models behind it.

CueCode

Cloud mode on Cue Cloud. Same models and pack include; CueCode can send denser agent signals for this stack.

Scripts & tools

Any OpenAI SDK or raw HTTP client works. Use it in CI or custom agents — the load we optimize for is still IDE coding.

Streaming

Agents need tokens as they land. Send stream: true on chat completions — standard SSE-shaped streaming for OpenAI-compatible clients.

stream sketch · deepseek-v4-pro

curl https://api.cuecloud.io/v1/chat/completions \
  -H "Authorization: Bearer cue_••••••••" \
  -H "Content-Type: application/json" \
  -N \
  -d '{
    "model": "deepseek-v4-pro",
    "stream": true,
    "messages": [
      { "role": "user", "content": "refactor this module" }
    ]
  }'

CueCode vs everyone else

API

Standard OpenAI-compatible semantics. Cursor, Claude Code, custom tools, and scripts all aim at the same gateway.

CueCode

Same models and pack include. CueCode can send richer agent signals for better density and latency on this stack.

Works anywhere. Tuned for coding IDEs.

Your cue_… key is not locked to one app — use it in scripts, CI, or custom tools. Cue Cloud is optimized for IDE coding agents: repo context, tool loops, multi-step fixes. Generic chat works; agentic coding in an IDE is where this stack is meant to shine.

Errors

Code	Meaning
`401`	Missing or invalid cue_… key
`400`	Bad request or unknown model id
`503`	Capacity path unavailable — retry / check status

Pricing

$500 / month = 1 pack = 1B tokens (in + out) this month. Packs →

Back to explorer