# GeiserX/lynxprompt-mcp [Health: Active]

**Category:** ☁️ Cloud Platforms  
**Repository:** https://github.com/GeiserX/lynxprompt-mcp  
**GitHub Stars:** 2  
**npm Downloads (last month):** 1202  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/geiserx-lynxprompt-mcp

## Description
Go-based MCP server for LynxPrompt prompt management. Create, search, and organize AI prompts and AGENTS.md/CLAUDE.md configuration files. Docker image available.

## Tools
Capabilities this server exposes over MCP:

- **create_blueprint** — Create a new blueprint in LynxPrompt
- **create_hierarchy** — Create a new hierarchy in LynxPrompt
- **delete_blueprint** — Delete a blueprint from LynxPrompt
- **delete_hierarchy** — Delete a hierarchy from LynxPrompt
- **search_blueprints** — Search the public LynxPrompt blueprint marketplace
- **update_blueprint** — Update an existing blueprint in LynxPrompt

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "lynxprompt-mcp": {
    "command": "npx",
    "args": ["-y","lynxprompt-mcp"],
    "env": {
      "LYNXPROMPT_TOKEN": ""
    }
  }
}
```

**Requires environment variables:** `LYNXPROMPT_TOKEN` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What GeiserX/lynxprompt-mcp does

GeiserX/lynxprompt-mcp gives MCP-compatible AI clients access to a LynxPrompt instance. It can search the public blueprint marketplace, list a user's blueprints and hierarchies, read blueprints the authenticated user can access, and organize blueprints into repository-based hierarchies. Write operations cover creating, updating, and permanently deleting blueprints, plus creating and deleting hierarchies.

The server works with the hosted LynxPrompt service at `lynxprompt.geiser.cloud` or with a self-hosted deployment. Blueprints can represent AI configuration files such as `AGENTS.md` and `CLAUDE.md`; publishing a blueprint is controlled through its visibility when it is created.

## How it works

Authentication uses an API token created in the LynxPrompt instance's Settings area. The server communicates with LynxPrompt on behalf of the MCP client and exposes both tools and read-only MCP resources. Available resource URI patterns include:

- `lynxprompt://blueprints`
- `lynxprompt://blueprint/{id}`
- `lynxprompt://hierarchies`
- `lynxprompt://hierarchy/{id}`
- `lynxprompt://user`

Blueprint updates support optimistic concurrency: `get_blueprint` returns a `content_checksum`, which can be supplied as `expected_checksum` to prevent an update from silently replacing a change made elsewhere. Hierarchies group blueprints by repository root; creating one for an existing `repository_root` returns the existing hierarchy.

The process supports stdio and stateless Streamable HTTP at `/mcp`. HTTP deployments include Origin checks, a request body limit, and bearer authentication when exposed. The binary supports MCP revisions from 2024-11-05 through 2026-07-28.

## Setup and configuration

Install the npm package with an MCP client's command configuration, or run the published Docker image in stdio mode. The npm launcher downloads the release binary on first use and verifies its SHA-256 checksum against checksums included in the package. Docker Compose, local builds, and HTTP deployment are also supported.

The required environment variable is:

- `LYNXPROMPT_TOKEN`: API token for the LynxPrompt instance.

For a self-hosted instance, set `LYNXPROMPT_URL` to the server URL. Set `LYNXPROMPT_READ_ONLY=true` to register only read tools and hide write operations. The repository includes configurations for Claude Code, Claude Desktop, Cursor, VS Code, and Codex, as well as a Docker stdio example.

## Tools and capabilities

Read operations include `search_blueprints`, `list_blueprints`, `get_blueprint`, `list_hierarchies`, and `get_hierarchy`. Marketplace search accepts filters such as query, configuration or command type, category, tags, sort order, and page.

Write operations include `create_blueprint`, `update_blueprint`, `delete_blueprint`, `create_hierarchy`, and `delete_hierarchy`. Blueprint deletion and hierarchy deletion are destructive; deleting a hierarchy keeps its blueprints but removes their association. Each tool provides metadata describing its title, read-only or destructive behavior, and output schema.

## Limitations and notes

A token is required for access. Read-only mode removes write tools rather than merely blocking their use after registration. Updates should use the checksum returned by `get_blueprint` when preserving concurrent edits matters. Deleting a blueprint is permanent, while deleting a hierarchy leaves its blueprints intact but unlinked.

_Full upstream README: https://allmcps.com/mcp/geiserx-lynxprompt-mcp/readme_