Use CoralBricks with Cline
Run Cline's agent on CoralBricks open-model inference — GLM and Kimi with up to 1M context. Three fields, no plan gates.
Cline is the most-installed open-source AI coding agent
for VS Code (and VS Code forks like Cursor), with first-class support for
OpenAI-compatible providers. CoralBricks plugs in with three fields — plan
mode, act mode, and tool calling all route through the Coral gateway.
One command to connect, then the agent loop — file edits, shell commands,
streaming — runs on GLM 5.2 through the Coral gateway.
1. Get an API key
Create a key from your API keys page (format cb_...).
2. Configure the provider
Open Cline → gear icon (Settings) → API Configuration:
- API Provider →
OpenAI Compatible
- Base URL →
https://inference.coralbricks.ai/v1
- API Key → your
cb_... key
- Model ID →
glm-5.2-fp4 (or any slug from the table below — IDs are
case-sensitive)
Optional but recommended: in the custom model configuration, set the
context window size to match the model (1,048,576 for GLM 5.2 and
Kimi K3) so Cline's context management uses the full window.
Available models
| Model |
Model ID |
Context |
Input $/M |
Output $/M |
| GLM 5.2 |
glm-5.2-fp4 |
1M |
$1.12 |
$4.40 |
| Kimi K3 |
kimi-k3 |
1M |
$3.00 |
$15.00 |
| GPT-OSS 120B |
gpt-oss-120b |
128K |
$0.12 |
$0.60 |
Cached input tokens are always free — and Cline's agent loop re-sends its
context on every turn, which is exactly where cached-input pricing matters.
Cline CLI
The terminal version of Cline (shown in the recording above) connects with a
single command:
cline auth openai -k cb_... -b https://inference.coralbricks.ai/v1 -m glm-5.2-fp4
cline "Create merge_intervals.py with a function that merges overlapping intervals, add a couple of asserts, and run it."
The openai provider id is Cline's built-in OpenAI-compatible client; the
setting persists, so subsequent cline invocations use CoralBricks without
flags.
Optional: the CoralBricks plugin (CLI / SDK)
It's optional because the cline auth openai setup above is already a
complete integration — everything works without installing anything extra.
The plugin's job is to remove the manual configuration: it registers
coralbricks as a first-class named provider, with the base URL, model
catalog, context windows, and pricing preconfigured. With it active you get:
cline -P coralbricks -m glm-5.2-fp4 — select CoralBricks by name, no
base URL or generic openai slot to fill in (and nothing to reconfigure
if you use that slot for another endpoint)- correct per-model context windows out of the box (1M for GLM 5.2 and
Kimi K3), so Cline's context management uses the full window instead of a
generic default
- accurate cost tracking in Cline's own usage display, priced from our
published rates — including free cached input
- auth from the environment (
CORAL_API_KEY), convenient for scripted or
CI use of the Cline SDK
cline plugin install npm:@coralbricks/cline-plugin
export CORAL_API_KEY=cb_...
Note: released Cline CLI versions do not yet activate provider-capability
plugins; the plugin installs cleanly today and lights up as soon as a CLI
release enables the providers runtime. Until then, use cline auth openai
as above.
Notes
- Everything routes through Coral — unlike Cursor, Cline has no
model backend of its own: plan mode, act mode, tool calls, and diffs all
run on the provider you configure, with no paid tier requirement.
- Tool calling works out of the box — the gateway emits OpenAI-exact
streaming tool calls; Cline's file edits, terminal commands, and browser
actions run unmodified.
- Long context is the point — GLM 5.2 and Kimi K3 take up to 1M tokens,
so large-repo context fits without aggressive truncation.
- The same key works in OpenCode, Cursor,
Continue, and anything else that speaks OpenAI-compatible providers. See
the API reference for the full surface.