Kilo Code

Kilo Code is an open-source AI coding agent that runs inside VS Code, JetBrains IDEs (IntelliJ, PyCharm, WebStorm), as a CLI, in the cloud, and via Slack. Started as a fork of Roo Code and built by Kilo-Org on $8M in seed funding, Kilo was acquired by Anaconda in July 2026 and continues as an open-source product for individuals, teams, and organizations. It has 26.9k GitHub stars and 3M+ users processing over 40 trillion tokens. Trusted by developers at Meta, Amazon, Airbnb, PayPal, Square, and Red Hat. The product rebranded from kilocode.ai to kilo.ai in 2026 and won Product Hunt's #1 Open Source Product of the Month. The agent operates across five built-in modes: Architect for system planning, Code for implementation, Debug for diagnosis, Ask for safe codebase exploration, and Orchestrator for multi-step task management with parallel subagent delegation across separate git worktrees. Cloud Agents support headless background execution. A dedicated Security Agent and Code Reviewer extend the platform beyond code generation into security analysis and pull request review. Integration depth spans the full developer workflow: VS Code, JetBrains (IntelliJ, PyCharm, WebStorm), CLI, Cloud Agents, and Slack for team-based task assignment. The platform supports 500+ models via BYOK through OpenRouter, Anthropic, OpenAI, Google Gemini, and any compatible API endpoint with no markup. MCP server integration enables custom tool connections. Gateway and Gastown provide additional infrastructure surfaces. KiloClaw offers managed OpenClaw deployments for teams wanting hosted agent orchestration. Kilo separates pricing into three independent parts: the platform plan, AI inference, and cloud compute. The platform plan is free and open source forever for individuals under an MIT license. Teams is $15 per user each month for centralized billing, usage analytics, and shared agent management, and Enterprise is custom-priced for SSO, SCIM, audit logs, and SLAs. AI inference runs free through BYOK or local models at no subscription, pay-as-you-go through the Kilo Gateway at exact provider rates with no markup, or on an optional Kilo Pass subscription from $19 per month. Kilo Pass costs the same in either billing cycle and only the credit allowance moves, with an annual commitment earning 50 percent bonus credits against up to 40 percent month to month. Kilo states that from September 1, 2026 at the earliest it will add a 5 percent processing fee on card purchases of credits and begin billing for awake cloud compute time. That fee is stated not to apply to Teams and Enterprise seats, or to payment by invoice, wire, or ACH. KiloClaw managed OpenClaw deployments are a separate product at $55 per month month-to-month or $51 per month on a six-month commit, with AI inference billed separately. Kilo maintains a SOC 2 Type I report through its Trust Center. Kilo Code is not the right fit for teams that need mature enterprise governance out of the box. GitHub Copilot and Cursor ($16/mo billed annually) provide more polished SSO, audit logging, and RBAC for procurement-heavy enterprise organizations. Tuning Memory Bank, MCP tools, and custom modes has a real learning curve that Cursor or Copilot do not require. Local model performance depends heavily on available hardware, which limits local-first deployments on standard developer machines. Current state Q3 2026: Kilo Code has grown to 26.9k GitHub stars and 3M+ users processing over 40 trillion tokens. The platform expanded from IDE and CLI to include Cloud Agents, Slack integration, Security Agent, Code Reviewer, Gateway, and Gastown. KiloClaw launched for managed OpenClaw deployments. The product won Product Hunt's #1 Open Source Product of the Month, and in July 2026 Anaconda acquired Kilo, integrating it as the agentic engineering layer of its AI development platform. Kilo routes work to third-party model providers, some of which may use prompts for training, and ships an organization-level code training opt-out alongside provider filtering by data policy. Fully local model operation is supported for sensitive codebases.