Kilo Code

Carry the same coding-agent workflow across VS Code, JetBrains, and the CLI

The official README positions Kilo Code as an open-source coding agent for VS Code, JetBrains, and the CLI. It is not confined to one editor surface: the same product family can work against a codebase from an IDE or terminal, connecting interactive development with scriptable workflows.

Sources: [1]

Switch among Code, Plan, Ask, Debug, and Review agents for different stages of development

Kilo ships with Code, Plan, Ask, Debug, and Review agents. They cover multi-file code generation, architecture planning, codebase questions, troubleshooting, and review across performance, security, style, and test coverage. Users can also create custom agents, while self-checking, terminal command execution, and browser control extend the workflow beyond text generation.

Sources: [1]

Combine 500+ model choices with MCP extensions inside the agent workflow

The README says Kilo offers more than 500 models and allows switching models in the middle of a task. Kilo states that its model access uses provider pricing with zero markup and does not require individual API keys to get started. An MCP marketplace can also connect MCP servers, bringing model choice and tool extension into the same agent environment.

Sources: [1]

Use `kilo run --auto` to move from interactive work into autonomous CI/CD execution

The CLI supports kilo run --auto for fully autonomous operation without permission prompts. The README presents this as a CI/CD-oriented mode, including an example where the agent runs tests and fixes failures. This lets teams choose between human-confirmed IDE/CLI work and unattended pipeline execution using the same CLI.

Sources: [1]

Best suited to developers who want one agent across IDEs and terminals without fixing one model or role

Kilo is a strong fit for developers who want the same agent workflow in VS Code or JetBrains and in a terminal, teams that switch between planning, debugging, and review roles, and users who want to compare models by latency, cost, or reasoning during a task. It is also relevant when MCP extensions and CI automation need to be evaluated as part of one coding-agent stack.

Treat autonomous permissions and hosted Kilo surfaces as separate security boundaries

The README explicitly warns that --auto disables every permission prompt and lets the agent execute actions without confirmation, so it should only be used in trusted environments. CI deployments should still minimize repository write access, secrets, network access, and executable commands at the runner level.

Kilo also offers hosted Cloud Agent, automated Code Reviews, and KiloClaw alongside its open-source VS Code, JetBrains, and CLI clients. Local client use and hosted services are not the same deployment boundary, so organizations should evaluate each surface and its model/provider path separately. The CLI is a fork of OpenCode enhanced for the Kilo platform. The code is MIT licensed.

Sources: [1][2]