Kimchi

Kimchi is a terminal coding agent and inference layer that provides hard budget caps and spend attribution.

Kimchi is a governed AI coding platform combining a CLI coding agent, inference layer, and spend console. It allows engineering teams to use open-source models while maintaining control over spending, data, and deployment. The platform enforces hard budget caps and real-time spend attribution by user, team, API key, or model. Prompts, code, and outputs are excluded from model training to ensure data privacy. It can be deployed in Kubernetes on AWS, GCP, Azure, on-premises, or in your VPC.

Kimchi at a glance

Best for: Software engineer
Runs on: CLI, Web
Pricing: Open source free plan; Coder plan; Teams plan; Enterprise custom pricing
Released: April 2026
Category: AI coding tool
Website: kimchi.dev

Key features

CLI coding agent
Hard budget caps
Real-time spend attribution
Zero model-data egress
Multi-model orchestration
Team budget monitoring and RBAC
Self-hosted deployment option

Use it for

Plan, build, and review code using multiple models
Set hard spending limits by user, team, or API key
Deploy code generation infrastructure within a VPC
Perform AI-powered pull-request reviews on GitHub and GitLab

Companions

Tools that articles about Kimchi recommend using with it, each with its job.

OpenRouter — 300+ models on one key, with routing and failover rules.
Ollama — Local inference and free embeddings for codebase indexing.
LM Studio — OpenAI-compatible local endpoint Kilo can call.
JetBrains — Native plugin for IntelliJ, PyCharm, WebStorm and more.
Qdrant — Vector store for Kilo's opt-in semantic indexing.

Alternatives to Kimchi

Kilo Code — Open-source agent with any model and no markup on inference
OpenCode — Works with 75+ model providers, local models, and existing subscriptions
Freebuff — Access multiple coding AI models for free without subscriptions.
Command Code — Repairs open model tool calls and learns developer patterns automatically.
Kimi Code — Coding agent on the open-weight Kimi K3 model with a 1M context
SmallCode — Runs fully local with budget-managed context and a forgiving parser for small models.
Anti Samcheonpo — Catches redundant AI tasks and blocks wasted agent spending.
codemax — Combines multiple paid AI coding plans into a single terminal to bypass usage limits.
Coucou — It provides open-source, real-time agent monitoring and approvals directly from the screen notch.
Kiro — Spec-driven coding with property-based tests to catch hidden bugs
Dao Code — It uses DeepSeek's low cost and prefix caching to keep expenses extremely low.
Dirac — Executes complex coding tasks efficiently with hash-anchored edits and low token overhead.

All Kimchi alternatives