Neura Market
Kimchi.dev – managed AI endpoints that scale to your VPC, no migration logo

Kimchi.dev – managed AI endpoints that scale to your VPC, no migration

Free
FreeFree tier
Type
Open Source
Visit Website GitHub (2 stars)

About Kimchi.dev – managed AI endpoints that scale to your VPC, no migration

Kimchi Coding is a complete AI-powered coding platform that combines a CLI coding agent, an inference layer, and a spend console into a single open-source tool. It leverages frontier open‑source models (e.g., Kimi‑K2.7, Minimax‑M3, Nemotron‑3) with automatic multi‑model orchestration that routes tasks to the best model per job type and cost. Key workflows include /ferment, which autonomously breaks tasks into milestones and delivers a finished pull request without babysitting, and /code for interactive, conversational code sessions. The platform offers one‑command migration from Claude Code (preserving MCP servers and skills), an evaluator that grades each milestone for regressions and edge cases, and /teleport for persistent remote workspaces. A free Community tier provides generous rate limits, while paid plans add priority speed, larger context windows, remote workspaces, and self‑hosted VPC deployment for enterprises.

Key Features

CLI coding agent with interactive code sessions
Ferment workflow: prompt to autonomous pull request generation
Built-in multi-model orchestration routing tasks to best OSS model per type and cost
Evaluator that grades each milestone against original intent, catching regressions and edge cases
One-command setup with automatic migration from Claude Code and OpenCode
Teleport remote workspaces: persistent sandboxes pre-loaded with tools and skills
Self-hosted VPC deployment option (Enterprise) for data sovereignty
Spend console for usage and cost monitoring
Open-source models with no rate limits or token caps on community tier

Pros & Cons

Pros
  • Free tier with generous rate limits and no credit card required
  • No token caps or rate limits on the Community tier
  • One-command migration from Claude Code preserving MCP servers and skills
  • Multi-model orchestration reduces costs by using cheaper OSS models where appropriate
  • Self-hosted VPC option for enterprises needing data sovereignty
  • Open-source transparent codebase
  • Evaluator ensures quality and catches regressions before PR submission
Cons
  • Community tier limited to 250k context window and shared queue request speed
  • Teleport remote workspaces limited to 1 for Coder plan, 3 for Teams (extra $15/mo)
  • Some features (Kimchi Studio, Live Remote Agents) are listed as 'coming soon'
  • Enterprise pricing required for self-hosted VPC deployment

Best For

Automated code generation and refactoring via natural language promptsAutonomous pull request creation with milestone-based planning and evaluationInteractive code sessions for fine-grained control over agent behaviorRemote development with persistent workspaces that run in the cloudTeam collaboration with shared kanban boards (Kimchi Studio) and budget monitoring

FAQ

Which models can I run with Kimchi?
Kimchi uses a multi-model harness that stacks the strongest open source and open-weight models, including Kimi-K2.7, Minimax-M3, and Nemotron-3, with automatic orchestration by task type.
Can I switch plans anytime?
Yes, you can upgrade or downgrade at any time and changes apply right away.
What payment methods do you accept?
Kimchi accepts major credit cards and integrates with all major cloud provider marketplaces. Enterprise customers can request invoicing.
What happens if I hit rate limits?
The Community tier runs on a shared queue with generous rate limits. Coder and Teams tiers include priority request speed.