Reika

Reika is a coding agent command line interface designed for local and hosted language models.

Reika is a coding agent command line interface designed around small local models first, such as 8B to 35B parameter models. It helps developers work with local and hosted models by reducing context waste and handling repetitive read loops. It features context management that collapses old tool outputs and lets models write findings notes before turns fold. It provides safety controls, including approval prompts for dangerous commands and a macOS kernel sandbox that confines writes to the project. Users can run multiple modes like planning, vibe, minimal, grind, chat, and shell. It supports OpenAI-compatible model servers such as llama.cpp, MLX, vLLM, and cloud application programming interfaces.

Reika at a glance

Best for: Software engineer
Runs on: macOS, Linux, Windows, CLI
Pricing: Free
Released: October 2026
Category: AI coding tool
Website: github.com/alexwkleung/reika

Key features

Context discipline with collapsed tool outputs
Loop and spiral breaking with escalating ladders
Typechecking edits against pre-edit baselines
Read-only plan mode and vibe mode
Kernel sandboxing for model-chosen shell commands on macOS
Extensible via Model Context Protocol servers and markdown skills
Status bar displaying decode speed and context fill

Use it for

Writing and editing code with local or hosted models
Exploring codebases in a read-only plan mode
Running automated coding agent procedures
Executing shell commands with output integration

Alternatives to Reika

RTK — Reduces AI context output noise by 56 to 90 percent with zero configuration.
jevgrep — Reduces coding agent costs by providing targeted, relevant repository context.
Command Code — Repairs open model tool calls and learns developer patterns automatically.
OpenCode — Works with 75+ model providers, local models, and existing subscriptions
Ret — Manage multiple AI coding agents across projects from one unified desktop interface.
SmallCode — Runs fully local with budget-managed context and a forgiving parser for small models.
Orca — Runs multiple AI coding agents simultaneously in isolated workspaces.
fast-agent — Improves agent accuracy and cost efficiency via MCP integration.
Mdt — Avoids unnecessary rereading, token use, and distraction for agents.
Invariant — Persistent isolated Linux environments and parallel coding agent experiments.
Context7 — Provides up-to-date documentation for AI agents and code editors.
Wiggum CLI — Automates development cycles from backlog issues to merged pull requests.

All Reika alternatives