Kimi Code

Kimi Code CLI is Moonshot AI’s terminal coding agent. Register ApiGo as a Kimi provider so Kimi Code uses your ApiGo key, the kimi-k3 model, Workspace limits, and call logs.

Prerequisites

Kimi Code CLI installed
An ApiGo API key from Workspace → API Keys
kimi-k3 available in the Model Catalog

1. Install Kimi Code

On macOS or Linux, use the official installer:

curl -fsSL https://code.kimi.com/kimi-code/install.sh | bash

On Windows PowerShell:

irm https://code.kimi.com/kimi-code/install.ps1 | iex

Open a new terminal and confirm the command is available:

kimi --version

2. Configure ApiGo

Create or edit ~/.kimi-code/config.toml. If the file already contains other providers or models, merge the following entries instead of replacing them.

default_model = "kimi-k3"
default_permission_mode = "auto"
default_plan_mode = false
merge_all_available_skills = true
telemetry = true

[providers."managed:apigo"]
type = "kimi"
base_url = "https://api.apigo.ai/v1"
api_key = "sk-apg-..."

[models."kimi-k3"]
provider = "managed:apigo"
model = "kimi-k3"
max_context_size = 1048576
capabilities = [ "thinking", "always_thinking", "image_in", "video_in", "tool_use" ]
display_name = "K3"
support_efforts = [ "max" ]
default_effort = "max"

This configuration has been verified with ApiGo. The provider name managed:apigo is a local alias; the model entry must reference the same value.

Kimi Code stores api_key as plain text in this file. Restrict access after saving it:

chmod 700 ~/.kimi-code
chmod 600 ~/.kimi-code/config.toml

default_permission_mode = "auto" lets the Agent run tool actions without asking for confirmation. Change it to manual if you want to approve actions individually.

3. Start and verify

Start Kimi Code inside your project:

cd /path/to/your/project
kimi

Use /model to confirm that K3 is selected. Then send a short prompt, or run a one-shot terminal check:

kimi -p 'Reply with exactly APIGO_KIMI_OK and nothing else.'

A successful setup prints:

APIGO_KIMI_OK

Confirm the request in Workspace → Call Logs.

Configuration reference

Setting	Value	Purpose
Configuration file	`~/.kimi-code/config.toml`	Stores Kimi Code runtime, provider, and model settings
Provider type	`kimi`	Uses Kimi Code’s OpenAI-compatible provider implementation
Provider alias	`managed:apigo`	Connects the model entry to the ApiGo provider
Base URL	`https://api.apigo.ai/v1`	ApiGo OpenAI-compatible gateway prefix
Model alias and slug	`kimi-k3`	Selects the ApiGo K3 model
Context size	`1048576`	Declares the model’s 1M-token context window
Thinking effort	`max`	Uses the supported maximum thinking effort

Troubleshooting

Symptom	Fix
Configuration warning on startup	Run `kimi doctor` and confirm the file is valid TOML
401 / 403	Replace `api_key` with an active ApiGo key and check its Workspace permissions or spend cap
Model not found	Confirm both model fields are exactly `kimi-k3` and check the current ApiGo Model Catalog
Kimi Code ignores the file	Check whether `KIMI_CODE_HOME` points to another data directory
Kimi Code asks for tool approval	Confirm `default_permission_mode` is `auto`; use `manual` when approval prompts are preferred
No request in Call Logs	Confirm the provider uses `https://api.apigo.ai/v1` and `/model` shows K3

Resources

Hermes Agent

OpenClaw