Kilo Code — The Open-Source All-in-One Coding Agent
Build, ship, and iterate faster with AI in VS Code, JetBrains, or the CLI. 500+ models, zero markup.
- Source repo
- Kilo-Org/kilocode
- Stars
- ★ 28k
- Last updated
- 1d ago
- License
- MIT
- Primary language
- TypeScript
- FA score
- 55/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platform
- Cost
- Free tier plus a paid hosted plan
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- Quickly implement multi-file features in VS Code using natural language.
- Not a fit if
- Teams that need fully offline use or data that never leaves the machine
- Users unwilling to create an account or send code to a cloud service
- Teams needing a mature toolchain with complete documentation
- Source review
- 55/100 · Major gaps
What does this agent do, and when should you use it?
Kilo Code is an open-source AI coding agent available as a VS Code extension, JetBrains plugin, and CLI tool. It lets users choose from 500+ models and switch mid-task, paying only the provider's rate with no markup. No API keys are required to start. The repository includes the CLI (a fork of OpenCode), VS Code extension, JetBrains plugin, and supports code generation, inline autocomplete, self-checking, terminal and browser control, and an MCP marketplace. Cloud agent and code review services are also offered.
Kilo Code takes natural language instructions in the IDE or terminal, generates and edits code across multiple files, provides ghost-text autocomplete with tab-to-accept, and self-checks its own output. It can run terminal commands, automate the web via browser control, and integrate external MCP servers from the marketplace. Users can switch between models mid-task to balance latency and cost. The CLI offers kilo run --auto for fully autonomous CI/CD operation without prompts.
- Quickly implement multi-file features in VS Code using natural language.
- Keep your JetBrains IDE workflow with AI-assisted coding in IntelliJ or other IDE.
- Automate test fixes in CI/CD pipelines with
kilo run --auto. - Select a specific model for complex reasoning, e.g., Claude Opus for architecture design.
- Get AI feedback on pull requests via the code review feature.
- Use Kilo from the browser without local setup via the cloud agent.
How do you install or deploy this agent?
VS Code: Install the Kilo Code extension from the VS Code Marketplace. JetBrains: Search 'Kilo Code' in the JetBrains Marketplace or in Settings → Plugins. CLI: Install via npm npm install -g @kilocode/cli or curl curl -fsSL https://kilo.ai/cli/install | bash, also available via pnpm, bun, Homebrew, or Arch Linux AUR.
How do you use this agent?
After installing, create an account to access 500+ models. For CLI, run kilo in any project directory to start an interactive session. Use kilo run --auto "run tests and fix any failures" for CI/CD autonomous runs. In the IDE, open the extension panel to start coding.
What are this agent's strengths and limitations?
- Extensive model selection (500+) with mid-task switching at provider pricing
- Multi-platform support (VS Code, JetBrains, CLI, cloud) with a unified experience
- Built-in specialized agents (Plan, Debug, Review) optimized for different tasks
- MIT open source, free to modify and use commercially
- Autonomous mode for CI/CD integration
- Requires account creation for model access, raising privacy concerns
- CLI is a fork, potentially inheriting upstream limitations
- Huge model catalog may cause decision overload
- Autonomous mode requires trusted environments to avoid risks
- Documentation not fully covering configuration; external docs needed
How does this agent compare with similar options?
The source does not mention specific competitors.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Kilo Code — The Open-Source All-in-One Coding Agent This agent | 55 · Major gaps | IDE extensionFreemium | ★ 28k | 1d ago | TypeScript | — |
| Plannotator | 78 · Good | CLIFreemium | ★ 9.5k | today | TypeScript | Codex · Claude Code |
| Cline | 54 · Major gaps | CLIFree + model costs | ★ 71k | today | TypeScript | OpenAI API · Claude API |
| Continue | 43 · Major gaps | CLIFree + model costs | ★ 36k | 3mo ago | TypeScript | — |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: permission system exists but explicitly not a sandbox, only UX; --auto disables all prompts; SECURITY.md defines threat model but lacks data flow transparency; dependencies have overrides and patches but no vulnerability scan evidence; external effects include terminal and browser control but no rollback; source attribution via LICENSE and README fork statement. Deductions: permission system not security isolation, data flow not transparent, sensitive data handling not explicit, rollback missing.
Evidence shows: package structure clear, tests exist but not executed; dependencies pinned with overrides and patches; failure messages not evident in source. Deductions: static review cannot verify tests, failure messages not provided.
Evidence shows: multi-platform (VS Code, JetBrains, CLI), multiple agents and custom agents; capability boundaries listed in README; trigger precision not explicit; environment fit with install instructions. Deductions: trigger precision not explicit.
Evidence shows: README structure clear, install notes detailed, naming stable, FAQ and known limitations (SECURITY.md), MIT license, version exists but no changelog, maintenance responsibility via CONTRIBUTING and RELEASING. Deductions: no changelog.
Evidence shows: high output usability (multi-platform), high marginal value (500+ models), good cost-benefit (zero markup). Deductions: static review cannot verify actual effectiveness.
Evidence shows: README claims partially supported (e.g., fork), but lack independent verification; tests exist but not executed; facts and inferences not clearly separated. Deductions: lack independent verification, facts and inferences not separated.
- Permission system is not a security sandbox; --auto disables all prompts, use only in trusted environments.
- Data flow transparency is insufficient; users should understand data sent to LLM providers.
- No rollback mechanism provided; operations may be irreversible.
- Dependency security lacks vulnerability scan evidence; users should review dependencies themselves.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
jetbrains/v7.1.11 JetBrains 7.1.11
Release notes excerpt
Added - List remote agents for the mobile app. - Bundle Kilo CLI v7.8.13. Fixed - Prevent editor context ignore-file loading from blocking the JetBrains UI thread. - Guard deferred prompt dispatch. - Keep worker console writes off the TUI frame. - Refresh expired MCP OAuth tokens automatically. - Preserve session status occurrence timing. - Persist late OpenRouter reasoning details for replay. - Make OpenAI-compatible tool content normalization opt-in. Changed - Bump the Gradle wrapper to 9.8.1 and openapi-generator to 7.26.0.View on GitHub ↗ -
jetbrains/v7.1.10 JetBrains 7.1.10
Release notes excerpt
Added - Show and edit agent models from the agents list. - Configure transcript display, with a live preview and localized settings. - Reuse the Core submenu in the prompt bar menu. - Opt in to the Marketplace error handler. - Fork a remote session from a message in the CLI. - Let glob search an opt-in allowlist of ignored paths. - Bundle Kilo CLI v7.8.12. Fixed - Restore session lists after a CLI reconnect. - Stop overriding the user permission config. - Clear MCP sign-in state when the server is removed. - Make the model reset forget the saved per-mode pick. …View on GitHub ↗