Log inSubmit server
CodeZaiku logo

CodeZaiku

MCP coding, review, research, security, and SRE tools for local or hosted OpenAI-compatible models.

InstallRepositoryWebsite
Full README Compare alternatives Embed widget Claim this listing

Install CodeZaiku

Pick your client, then add it in one click or paste the config.

STDIONode.js
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "codezaiku": {
      "command": "npx",
      "args": [
        "-y",
        "@wyrdsekai/codezaiku-mcp"
      ]
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

Install not yet sandbox-tested— the server itself is confirmed live (we called its tools/list endpoint). The stdio install command is a separate, ongoing check.

Overview

CodeZaiku MCP server connects coding-agent workflows to local models or hosted OpenAI-compatible APIs. Its tools can edit and test projects, review code, investigate or repair Docker and host issues, research web sources, inspect security posture, and preserve project conventions. Reach for it when an agent needs a terminal-oriented backend with evidence-based checks, guarded changes, verification, and rollback behavior. It runs standalone and can also operate as a CLI subprocess or ACP agent.

Use cases

  • Fix and test a bug in an existing repository
  • Review a git diff for correctness and security issues
  • Investigate or repair an unhealthy Docker Compose stack
  • Research a technical question with cited web sources
  • Record project conventions for future coding tasks

Key features

  • Local or hosted OpenAI-compatible model support
  • Project editing with toolchain and test verification
  • Guarded SRE repair with rollback
  • Read-only code, incident, and security analysis
  • Cited web research with reusable memory
  • Persistent project coding conventions

Tools11

What CodeZaiku lets your agent do. Expand a tool for its parameters.

Verified live~2.1k tokens of context

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked 1d ago

Recent health check succeeded.

9/9 checks healthy over the last 37d

What it can do

From the 11 tools this server returned when we ran it.

Handles credentials · 1Makes network requests · 1Read-only tools · 11
Network at startup5 connections
Files written at startup1
Install scriptsNone
npm provenanceSigned build
Package age40 days
Maintainers1

Observed 2026-10-22 in an isolated sandbox. A description of reach, not a security audit.

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 14d ago via OSV.dev · @wyrdsekai/codezaiku-mcp (npm)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars1Stargazers on the source repository.
npm downloads1.8kPackage downloads in the last 30 days.
Last commit6d agoMost recent push to the default branch.
Tools exposed11Callable tools this server registers over MCP.
Directory activity1 viewsConfig copies, upvotes, and views on AllMCPs.

How CodeZaiku works

What CodeZaiku does

CodeZaiku provides an MCP interface for software development and operational work. It uses an OpenAI-compatible model server, either on the local machine or through a hosted API, and does not include model weights. The reference setup targets a 9B model, but the harness can connect to compatible model servers such as Ollama, llama.cpp, LM Studio, vLLM, or other OpenAI-compatible endpoints.

Its coding workflow reads an existing project, changes files, runs the project toolchain, and checks the result with the project's own tests. Separate read-only tools cover code review, incident diagnosis, security assessment, and web research. Operational tools can inspect or repair Docker Compose stacks and hosts, while convention tools maintain project-specific coding guidance for later coding jobs.

How it works

The code tool blocks while it completes a coding task such as a bug fix, feature, refactor, test-coverage change, incident task, or API migration. review examines a project or git diff without modifying it and returns findings with severity, file and line references, problems, and suggested fixes.

investigate diagnoses a stack or host without remediation. fix can apply and verify a bounded repair, roll back a failed change, and apply an authority ladder ranging from observation through unattended action. explore_and_fix broadens that process to a whole machine by finding Docker Compose stacks, failed systemd units, and host disk or memory pressure. It runs the operator on unhealthy Compose stacks and surfaces system or host problems.

secure reports container and host security concerns, including privileged containers, Docker socket mounts, publicly exposed datastores, weak credentials, image CVEs, and runtime intrusion detections. For a live detection, it proposes one read-only containment command but does not execute containment. research searches and reads web sources, then returns a cited synthesis; broad mode surveys sources, while depth mode focuses on one question. research_memory recalls findings accumulated from earlier research runs.

Setup and configuration

CodeZaiku requires an OpenAI-compatible model server. A JDK 21 or newer is required unless using a distribution that includes its own Java runtime. Linux is the reference platform; coding and research surfaces also run on macOS, remote Linux hosts can be operated over SSH, and Windows is supported through WSL2. A container image is available.

Installation options include the remote installer, a release tarball, Docker, a Debian package, or building from source. The installer places files under ~/.local on Unix-like systems by default and supports CODEZAIKU_PREFIX for a different destination. The command-line setup flow can detect model services, select an endpoint, register hosted models and keys, and run environment and connectivity checks.

CodeZaiku can be invoked as an MCP server with codezaiku mcp, as an ACP agent with codezaiku acp, or through the CLI subprocess contract with codezaiku run. The job_status tool polls asynchronous jobs submitted with async:true.

Tools and capabilities

  • code: edit, run, and verify project changes.
  • fix: diagnose, repair, verify, and roll back bounded operational changes.
  • explore_and_fix: triage a complete machine and its Compose stacks.
  • review: produce read-only correctness, security, and maintainability findings.
  • research and research_memory: research cited web sources and reuse stored findings.
  • investigate and secure: diagnose incidents and report security posture without remediation.
  • record_convention and show_conventions: maintain project coding conventions.
  • job_status: retrieve results from asynchronous tool jobs.

Read the full README View source on GitHub

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about CodeZaiku

CodeZaiku is an MCP server that connects coding agents to local or hosted OpenAI-compatible model APIs. Its main tools edit and test projects, review code, research cited web sources, investigate or repair Docker and host issues, assess security, maintain project conventions, and poll asynchronous jobs.

Install it with the provided remote installer, release tarball, Docker image, Debian package, or source build. A JDK 21+ is needed unless using a distribution with its own runtime. Then invoke the MCP interface with `codezaiku mcp`.

CodeZaiku works with OpenAI-compatible model servers, including Ollama, llama.cpp, LM Studio, vLLM, and hosted compatible APIs. It ships no model weights and can detect or configure model endpoints through its setup commands.

Yes. The `fix` tool can diagnose and repair a specified stack with blast-radius and authority controls, verification, and rollback on failure. `explore_and_fix` can scan a whole machine, find Compose stacks, and operate on unhealthy ones.

Yes. CodeZaiku exposes an MCP server interface, so it can be used with MCP-capable clients such as Cursor. Start the server through `codezaiku mcp` and configure the client to connect to it.

Related MCP servers

All in Coding Agents All alternatives
Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewCodeZaiku AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/codezaiku?style=directory)](https://allmcps.com/mcp/codezaiku)
HTML Embed
<a href="https://allmcps.com/mcp/codezaiku"><img src="https://allmcps.com/api/badge/codezaiku?style=directory" alt="CodeZaiku on AllMCPs" /></a>