mikuproject-mcp
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@mikuproject-mcplist available mikuproject tools"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
mikuproject-mcp
mikuproject-mcp is an MCP server adapter for mikuproject.
It exposes mikuproject operations to MCP clients as tools, resources, and
prompts. The primary entrypoint is local stdio. A localhost-oriented Streamable
HTTP entrypoint is also available for clients that need HTTP transport. The
server does not call external AI services and does not replace the upstream
mikuproject product logic.
Status
This repository is currently a Node.js / TypeScript MCP server implementation. The Node package metadata is prepared for future npm registry publication, but the current validation path is still to build and run it from a local checkout or use a GitHub Release tarball when one is available.
Requirements
Node.js 20 or later
npm
Bundled or configured
mikuprojectruntime artifacts
The checked-in runtime artifacts under runtime/ are used by the MCP adapter
for local execution.
Build
npm install
npm run buildThe stdio MCP server entrypoint after build is:
packages/node/dist/index.jsThe Streamable HTTP entrypoint after build is:
packages/node/dist/http.jsThe package name prepared for the Node release is:
@igapyon/mikuproject-mcp-nodeVersion Policy
The MCP package version should generally match the bundled Node.js
mikuproject CLI runtime version.
If the bundled Java runtime has a different patch version, treat that as runtime traceability information. Do not use the Java runtime patch version as the MCP package version unless the release is intentionally based on a Java-only runtime surface.
MCP Client Configuration
Use the built stdio entrypoint from your MCP client configuration.
Example:
{
"mcpServers": {
"mikuproject": {
"command": "node",
"args": ["packages/node/dist/index.js"]
}
}
}If your MCP client runs from a different working directory, use paths that are valid for that client environment.
Streamable HTTP Usage
The HTTP entrypoint is intended for local or otherwise explicitly controlled
deployments. By default it binds to 127.0.0.1:3000 and exposes a single MCP
endpoint at /mcp.
node packages/node/dist/http.jsYou can configure the listener with environment variables:
MIKUPROJECT_MCP_HTTP_HOST=127.0.0.1
MIKUPROJECT_MCP_HTTP_PORT=3000
MIKUPROJECT_MCP_HTTP_ENDPOINT=/mcp
MIKUPROJECT_MCP_HTTP_MAX_BODY_BYTES=52428800
MIKUPROJECT_MCP_HTTP_MAX_RESPONSE_BYTES=52428800
MIKUPROJECT_MCP_HTTP_ALLOWED_ORIGINS=http://localhost:3000The current HTTP transport is stateless. It does not issue Mcp-Session-Id,
does not keep durable project state, and creates a fresh MCP server instance for
each HTTP request. HTTP tool calls do not accept client-provided host file path
arguments such as statePath, workbookPath, inputPath, or outputPath.
Use inline content fields such as draftContent, workbookContent,
stateContent, patchContent, content, or inputBase64, and use
outputMode: "content" or outputMode: "base64" when the result should be
returned directly.
The HTTP entrypoint still creates a request-scoped temporary workspace for
adapter-internal runtime files when the upstream CLI requires one input as a
path. For example, a request can send stateContent and patchContent; the
adapter may materialize stateContent as a temporary runtime input while
passing patchContent on stdin. This workspace is removed after the response
completes and is not a durable storage location. Hosted or remote profiles that
need downloadable artifacts should add an explicit upload, download, or
content-return policy.
HTTP request bodies are limited by MIKUPROJECT_MCP_HTTP_MAX_BODY_BYTES.
HTTP response bodies are limited by MIKUPROJECT_MCP_HTTP_MAX_RESPONSE_BYTES,
which defaults to the same 50 MiB limit. If a generated content-mode or Base64
response would exceed the response limit, the HTTP entrypoint returns HTTP 413
with a JSON-RPC error. Use a future retained-download policy for large generated
artifacts.
The HTTP server rejects non-local Origin headers unless they are listed in
MIKUPROJECT_MCP_HTTP_ALLOWED_ORIGINS. Do not bind it to a public interface
without a separate authentication, workspace isolation, upload, storage,
cleanup, audit, and runtime isolation design.
Release Tarball Usage
GitHub Releases may provide an npm package tarball asset named like:
igapyon-mikuproject-mcp-node-0.8.3.tgzAfter downloading the tarball, install it globally:
npm install -g ./igapyon-mikuproject-mcp-node-0.8.3.tgzThen configure your MCP client to run the installed command:
{
"mcpServers": {
"mikuproject": {
"command": "mikuproject-mcp"
}
}
}You can also run the release tarball directly with npm exec without a global
install. Replace the version and URL with the release asset you want to use:
npm exec --yes --package=https://github.com/igapyon/mikuproject-mcp/releases/download/v0.8.3/igapyon-mikuproject-mcp-node-0.8.3.tgz -- mikuproject-mcpExample MCP client configuration:
{
"mcpServers": {
"mikuproject": {
"command": "npm",
"args": [
"exec",
"--yes",
"--package=https://github.com/igapyon/mikuproject-mcp/releases/download/v0.8.3/igapyon-mikuproject-mcp-node-0.8.3.tgz",
"--",
"mikuproject-mcp"
]
}
}
}Runtime Configuration
By default, runtime artifacts are resolved from runtime/.
You can override them with environment variables:
MIKUPROJECT_MCP_RUNTIME_NODE=/path/to/mikuproject.mjs
MIKUPROJECT_MCP_WORKSPACE=/path/to/workspaceMIKUPROJECT_MCP_WORKSPACE controls where generated state, projections,
exports, reports, summaries, and diagnostics are written. If it is not set, the
server uses workplace/ under this repository.
Bundled runtime discovery accepts versioned artifact names such as
runtime/mikuproject-node/mikuproject-0.8.3.3.mjs and
runtime/mikuproject-java/mikuproject-0.8.3.3.jar. When multiple versioned
artifacts are present, the newest numeric version is selected. Legacy names
(mikuproject.mjs and mikuproject.jar) remain supported as fallback names.
Path and Content I/O
Path mode remains the default. Tools accept existing *Path fields and write
default outputs under MIKUPROJECT_MCP_WORKSPACE when outputPath is omitted.
The bundled Node.js mikuproject 0.8.3.3 runtime also supports content mode for
operations whose schemas expose inline fields. Text inputs can use
draftContent, workbookContent, stateContent, patchContent, or
content; XLSX imports can use inputBase64. When the upstream runtime accepts
only one stdin input for a multi-input operation, the adapter can materialize
one inline input as an adapter-owned temporary file while keeping host file paths
out of the HTTP contract. Text outputs can set outputMode to content; binary
outputs can set outputMode to base64. In those modes, the generated artifact
is returned in the tool result as text or base64 with a mimeType, and no
default output file is created for that artifact.
Java runtime execution remains path-mode only until equivalent stdin/stdout and Base64 behavior is verified. When content mode is requested, the adapter selects the Node.js runtime and reports a runtime diagnostic if no compatible Node.js runtime is available.
Tools
The server exposes product-prefixed tools derived from the upstream CLI command tree.
Core state and AI workflow tools:
mikuproject_versionmikuproject_ai_specmikuproject_ai_detect_kindmikuproject_state_from_draftmikuproject_ai_export_project_overviewmikuproject_ai_export_bundlemikuproject_ai_export_task_editmikuproject_ai_export_phase_detailmikuproject_ai_validate_patchmikuproject_state_apply_patchmikuproject_state_diffmikuproject_state_summarize
Import, export, and report tools:
mikuproject_export_workbook_jsonmikuproject_export_xmlmikuproject_export_xlsxmikuproject_import_xlsxmikuproject_report_wbs_xlsxmikuproject_report_daily_svgmikuproject_report_weekly_svgmikuproject_report_monthly_calendar_svgmikuproject_report_allmikuproject_report_wbs_markdownmikuproject_report_mermaid
Resources
Common resource URIs include:
mikuproject://spec/ai-jsonmikuproject://state/currentmikuproject://state/{name}mikuproject://export/workbook-jsonmikuproject://export/project-xmlmikuproject://export/project-xlsxmikuproject://report/wbs-xlsxmikuproject://report/daily-svgmikuproject://report/weekly-svgmikuproject://report/monthly-calendar-svgmikuproject://report/allmikuproject://report/wbs-markdownmikuproject://report/mermaidmikuproject://summary/{operationId}mikuproject://diagnostics/{operationId}
Tool results include generated artifact paths and, where applicable,
product-specific resource URIs. Content-mode results can instead include
result-only artifact fields such as text, base64, and mimeType.
Prompts
The server provides small product-specific prompts:
mikuproject_create_project_draftmikuproject_revise_state_with_patchmikuproject_review_artifact_diagnostics
Prompt text refers to the AI specification resource instead of duplicating the full product specification.
Diagnostics and Outputs
Tool results are JSON text results with:
okoperationoperationIddiagnosticsgenerated artifact references
Operation summaries and diagnostics are saved under the workspace and can be read through:
mikuproject://summary/{operationId}mikuproject://diagnostics/{operationId}
Security Notes
This server is intended for trusted local use. It invokes local runtime artifacts and reads or writes local files based on tool arguments and workspace configuration.
Do not configure it with untrusted runtime artifacts or expose it as a hosted network service without a separate design for authentication, workspace isolation, upload handling, storage policy, cleanup, audit, and runtime isolation.
Developer Documentation
Developer-facing repository layout, implementation order, and contract notes are
in docs/development.md and docs/miku-soft-50-mcp-design-v20260501.md.
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Latest Blog Posts
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/igapyon/mikuproject-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server