Kilo Code

What it does Kilo Code is an open-source agentic coding extension for VS Code, JetBrains, and the CLI. It edits files, runs commands, manages multi-step tasks, and uses MCP tools — bring your own LLM API key and pay provider rates with zero platform markup.

Who it's for Developers and small platform teams who want a real agentic coding tool without per-seat SaaS lock-in, and who prefer to pay LLM providers directly rather than a wrapper. Also a clean fit for contractors and consultancies juggling multiple client API keys.

How platform engineers use it Install the VS Code or JetBrains extension. Configure an OpenRouter, Anthropic, or OpenAI key (or run against any of 500+ models). The agent reads/writes files in the workspace, runs shell commands with approval, and connects to MCP servers for tool use. For Terraform, Kubernetes, and CI work the same loop applies — explain the change, agent proposes a multi-file edit, you approve. The optional Kilo Pass adds bonus inference credits if you don't want to manage provider billing directly. KiloClaw spins up a managed OpenClaw agent on a dedicated Firecracker microVM for scheduled tasks and chat-channel integration.

Strengths

• Genuinely open source (kilo-org/kilocode on GitHub) — auditable, forkable, no vendor lock
• Pay LLM providers at cost, no markup — meaningful for high-volume agentic workloads
• 500+ model support via OpenRouter, plus direct provider connections and BYOK at the team tier
• KiloClaw managed VMs use Firecracker isolation with an independent security assessment published

Limitations

• Polish, docs, and onboarding lag commercial competitors (Cursor, Claude Code)
• Product line is fragmented: Kilo Code, Kilo Pass, KiloClaw, Teams, Enterprise — pricing is non-obvious
• No completed third-party compliance certifications yet (controls in place, no SOC 2 report on the trust center)

AI maturity AI-native. The product exists only because agentic coding became viable; without LLMs there is no Kilo Code. The differentiator is the open-source distribution model and pricing transparency, not novel AI architecture — under the hood it's the same model selection (Claude Sonnet, GPT-5, Gemini, etc.) wired into an extension.