Dashboard
Your usage at a glance — from the last 24 hours of call records and the prepaid ledger.
Sign in to see your dashboard.
Sign inToken spend
prepaid credits—
spent from the current grants (the free allowance is $0)
Total tokens
in + out—
last 24 hours
Requests
—
last 24 hours
Avg latency
—
completed calls, last 24 hours
Requests over time
Calls per hour (last 24 hours)
Token split
Input and output, by model (last 24 hours)
- Input —
- Output —
Connect your tool
Pick a client for its setup block. Issue a connection token in the API keys panel below and, while this page is open, Copy inserts the real token (the block shows it masked).
export KOTOBA_API_BASE="https://api.kotoba.cloud"
export KOTOBA_API_TOKEN="kc_pat_<your-token>"
export KOTOBA_MODEL="qwen3.8-flash-next-whitehacker"
# test the connection
claude -p "Reply with exactly: OK"Caveat: api.kotoba.cloud does not serve the Anthropic Messages wire (POST /v1/messages) yet — measured 405 on 2026-09-15. To use Claude Code, put a local proxy that translates Anthropic → OpenAI (LiteLLM, for example) in between and point ANTHROPIC_BASE_URL at it. See the Anthropic SDK page.
export KOTOBA_API_BASE="https://api.kotoba.cloud/v1"
export KOTOBA_API_TOKEN="kc_pat_<your-token>"
# test the connection
codex exec "Reply with exactly: OK"OpenAI-compatible — base URL includes /v1.
Base URL: https://api.kotoba.cloud/v1
API Key: kc_pat_<your-token>
Model ID: qwen3.8-flash-next-whitehackerProvider ID: kotobacloud (lowercase)
Base URL: https://api.kotoba.cloud/v1
API Key: kc_pat_<your-token>
Model ID: qwen3.8-flash-next-whitehacker~/.config/opencode/opencode.json:
{
"provider": {
"kotobacloud": {
"baseUrl": "https://api.kotoba.cloud/v1",
"apiKey": "kc_pat_<your-token>",
"models": {"qwen3.8-flash-next-whitehacker": {}}
}
}
}Reasoning models can 400 on temperature — a known caveat.
export OPENAI_API_BASE="https://api.kotoba.cloud/v1"
export OPENAI_API_KEY="kc_pat_<your-token>"
# the openai/ prefix stacks on our model id
aider --model openai/qwen3.8-flash-next-whitehacker# ~/.hermes/config.yaml (or a profile's config.yaml)
model:
provider: kotoba
default: qwen3.8-flash-next-whitehacker
max_tokens: 8192
providers:
kotoba:
api: ${KOTOBA_API_BASE}/v1
key_env: KOTOBA_API_TOKEN
transport: chat_completions
request_timeout_seconds: 900
models:
qwen3.8-flash-next-whitehacker:
context_length: 1000000
# ~/.hermes/.env
KOTOBA_API_BASE=https://api.kotoba.cloud
KOTOBA_API_TOKEN=kc_pat_<your-token>
hermes chat -Q --oneshot -q "Reply with exactly: OK"Auxiliary title generation spends one free request per turn (`auxiliary.title_generation.enabled: false` stops it). Long answers wait up to 14 minutes on the synchronous call, so request_timeout_seconds 900.
docker run -d -p 3000:8080 \
-e OPENAI_API_BASE_URL="https://api.kotoba.cloud/v1" \
-e OPENAI_API_KEY="kc_pat_<your-token>" \
ghcr.io/open-webui/open-webui:main
# then open http://localhost:3000 and pick the modelBase URL: https://api.kotoba.cloud/v1
API Key: kc_pat_<your-token>
Model: qwen3.8-flash-next-whitehackerNeeds Cursor Pro. Requests reach the API but Cursor may not render the reply — a Cursor-side limitation.
from openai import OpenAI
client = OpenAI(
base_url="https://api.kotoba.cloud/v1",
api_key="kc_pat_<your-token>",
)
r = client.chat.completions.create(
model="qwen3.8-flash-next-whitehacker",
messages=[{"role": "user", "content": "Reply with exactly: OK"}],
)API keys
Issue and revoke connection tokens. A token is shown once.
Use from a CLI / IDE
Issue a connection token for a local CLI / IDE agent. It is shown once.