Skip to content
/

LynxPrompt

lynxprompt-geiserx · GeiserX/LynxPrompt · ★ 41 · last commit 2026-05-25

Primitive shape 1 total
MCP tools 1
00

Summary

LynxPrompt — Summary

LynxPrompt is a self-hostable, federated platform for managing AI IDE configuration files — AGENTS.md, CLAUDE.md, .cursor/rules/, slash commands, and 30+ other formats — deployed via Docker Compose with a Next.js web UI, PostgreSQL database, and a CLI (lynxp). The platform functions as an internal marketplace where teams create "blueprints" (configuration templates), share them within an organization, and export them to any supported AI tool format with one click. An interactive wizard (web + CLI) auto-detects tech stack from GitHub/GitLab URLs, supports template variables and monorepo hierarchies, and handles SAML/OIDC/LDAP for enterprise SSO. Optional AI-assisted blueprint editing via the Anthropic API (opt-in, requires user's own API key) is the only cloud-bound feature. The VS Code extension, npm-distributed CLI (lynxp), GitHub Action, n8n node, Helm chart, and MCP server form a broad distribution ecosystem — making LynxPrompt the most "enterprise-ready" distribution mechanism among all ten batch frameworks. Unlike seed frameworks that define agent behavior, LynxPrompt's value is the institutional knowledge management layer: standardizing which CLAUDE.md files exist across teams, versioning them, and distributing them at scale.

01

Overview

LynxPrompt — Overview

Origin

Created by Sergio Fernández Rubio (GeiserX on GitHub). Licensed under GPL-3.0. Website: lynxprompt.com. Hosted instance available for those who prefer not to self-host.

Philosophy

LynxPrompt is a config management platform — it treats AI tool configuration files (CLAUDE.md, AGENTS.md, .cursorrules) as first-class organizational assets that need versioning, sharing, and standardization the same way code does.

From the README:

"Instead of manually writing configuration files for every project and every AI tool, use LynxPrompt to: Generate configs through an interactive wizard / Share blueprints through a private or federated marketplace / Standardize AI behavior across projects with reusable templates / Export to any supported format with one click."

"LynxPrompt is free and open-source. Self-host it for personal use, or deploy it within your organization to enforce coding standards, share institutional knowledge, and ensure consistent AI assistant behavior across your engineering teams."

Key Design Principles

  1. Self-hostable first — Docker Compose as the deployment primitive, not SaaS
  2. 30+ format support — write once, export to any AI tool
  3. Federated marketplace — teams can share blueprints within or across organizations
  4. All-or-nothing auth — SAML, OIDC, LDAP, OAuth, passkeys, magic link — full enterprise auth stack
  5. Optional AI — AI-assisted editing is opt-in and uses the user's own Anthropic key; the platform works fully without it

Scope

LynxPrompt is the only framework in this batch that explicitly targets enterprise teams — multi-user, SSO-enabled, team-scoped blueprint sharing, self-hosted infrastructure. All other frameworks are single-developer tools.

Ecosystem Components

  • lynxprompt-mcp — MCP server for browsing blueprints from AI assistants
  • lynxprompt-vscode — VS Code extension
  • homebrew-lynxprompt — Homebrew tap
  • n8n-nodes-lynxprompt — n8n community node for workflow automation
  • lynxprompt-action — GitHub Action for CI/CD integration
  • Helm chart — Kubernetes deployment
  • AUR package — Arch Linux
  • Snap package — Ubuntu/Snap
  • Chocolatey package — Windows
02

Architecture

LynxPrompt — Architecture

Distribution

  • Type: Self-hosted web app + CLI tool + VS Code extension + MCP server
  • CLI Binary: lynxp (also lynxprompt) from cli/dist/index.js
  • CLI Version analyzed: 2.1.0
  • Web App Version: 2.1.0
  • License: GPL-3.0
  • Runtime required: Docker (for server), Node.js (for CLI)

Install Methods

# Server (Docker Compose)
curl -O https://raw.githubusercontent.com/GeiserX/LynxPrompt/main/docker-compose.selfhost.yml
docker compose -f docker-compose.selfhost.yml up -d
# → http://localhost:3000

# CLI
npm install -g lynxprompt
# Point to self-hosted instance (optional):
lynxp config set-url https://lynxprompt.your-company.com

# Homebrew
brew install GeiserX/lynxprompt/lynxprompt

# Chocolatey (Windows)
choco install lynxprompt

# AUR
# (Arch Linux)
# Snap
snap install lynxprompt

# Helm (Kubernetes)
helm repo add lynxprompt https://geiserx.github.io/LynxPrompt
helm install lynxprompt lynxprompt/lynxprompt

Tech Stack

Layer Technology
Frontend + API Next.js 16 with App Router
Database PostgreSQL with Prisma ORM
Auth NextAuth.js (GitHub OAuth, Google OAuth, magic link, passkeys/WebAuthn, SAML, OIDC, LDAP)
Search PostgreSQL full-text search
Deployment Docker Compose with auto-migration on startup
Analytics Umami (optional)
Payments Stripe (optional, for marketplace)
AI Anthropic SDK (optional, for AI-assisted editing)

Directory Tree

LynxPrompt/
├── src/
│   ├── app/               # Next.js App Router pages
│   ├── components/        # React UI components
│   ├── lib/               # Auth, DB, utilities
│   ├── types/             # TypeScript types
│   └── proxy.ts           # API proxy
├── cli/
│   ├── src/
│   │   ├── index.ts       # CLI entry point
│   │   ├── config.ts      # CLI config
│   │   ├── api.ts         # API client
│   │   └── commands/      # CLI command implementations
│   │       ├── analyze.ts, check.ts, config.ts, convert.ts
│   │       ├── diff.ts, hierarchies.ts, import.ts, link.ts
│   │       ├── list.ts, login.ts, logout.ts, merge.ts
│   │       ├── pull.ts, push.ts, search.ts, status.ts
│   │       ├── whoami.ts, wizard.ts
│   └── package.json       # bin: {lynxprompt, lynxp}
├── prisma/                # DB schema + migrations
├── charts/lynxprompt/     # Helm chart
├── docker/                # Docker configs
├── action/                # GitHub Action
├── packages/              # Internal packages (MCP server, etc.)
└── docker-compose.selfhost.yml

Configuration Environment Variables

Key: NEXTAUTH_SECRET, APP_URL, ENABLE_GITHUB_OAUTH, ENABLE_EMAIL_AUTH, ENABLE_PASSKEYS, ENABLE_SSO, ENABLE_AI, AI_MODEL, ANTHROPIC_API_KEY, ENABLE_STRIPE, SUPERADMIN_EMAIL, ENABLE_USER_REGISTRATION

03

Components

LynxPrompt — Components

CLI Commands (18 subcommands)

Command Purpose
lynxp wizard Interactive step-by-step config generator
lynxp pull <bp_id> Pull a blueprint to local files
lynxp push Push local configs to server
lynxp search <query> Search blueprints
lynxp list List your blueprints
lynxp status Show status of local vs cloud configs
lynxp diff Diff local files against cloud versions
lynxp convert Convert between AI tool formats
lynxp merge Merge multiple blueprints
lynxp hierarchies Manage monorepo hierarchies
lynxp link Link a blueprint to a project
lynxp import <file> Import a blueprint file
lynxp analyze Analyze a config file
lynxp check Validate config files
lynxp login Authenticate (opens browser)
lynxp logout Log out
lynxp whoami Show current user
lynxp config View/edit CLI configuration

Web Platform Features

Feature Description
Blueprint Marketplace Internal/federated sharing of AI configs
Interactive Wizard Auto-detects stack from GitHub/GitLab URLs
Format Export One-click export to 30+ AI tool formats
Blueprint Variables Template variables in configs
Monorepo Support Hierarchical configs for subdirectories
Draft Autosave Autosaves wizard progress
REST API Full programmatic access (generate API tokens)
Admin Panel Superadmin promotion, user management
Blog Module Optional blog (ENABLE_BLOG)
Support Forum Optional forum (ENABLE_SUPPORT_FORUM)
Stripe Marketplace Optional paid blueprints (ENABLE_STRIPE)

Supported AI Tool Formats (30+)

Cursor, Claude Code, GitHub Copilot, Windsurf, Zed, Aider, Gemini CLI, Cline, Roo Code, Amazon Q, JetBrains Junie, and more.

Ecosystem Integrations

Tool Package Purpose
VS Code lynxprompt-vscode Browse/pull/diff configs in editor
MCP server lynxprompt-mcp Browse blueprints from AI assistants
GitHub Action lynxprompt-action CI/CD integration
n8n n8n-nodes-lynxprompt Workflow automation
Homebrew homebrew-lynxprompt macOS/Linux install
Helm charts/lynxprompt Kubernetes deployment
05

Prompts

LynxPrompt — Prompts

LynxPrompt is a config management platform, not a prompt framework. It stores and distributes AI tool configuration files (CLAUDE.md, AGENTS.md, .cursorrules) but does not ship its own agent prompts. The "blueprints" it manages ARE the prompts that users write for their AI tools.

Excerpt 1 — Interactive Wizard Output Pattern

The CLI wizard (lynxp wizard) generates config files. A representative output for a Next.js TypeScript project:

# Project Configuration

**Stack**: Next.js 15, TypeScript, PostgreSQL (Prisma), Tailwind CSS, Vitest

## Code Style
- TypeScript strict mode
- Functional components with hooks
- Server Components by default, Client Components only when needed
- Async/await over .then()

## Testing
- Vitest for unit tests
- Playwright for E2E
- Test files in __tests__/ or *.test.ts alongside source

## Key Conventions
- [conventions auto-detected from repo structure]

Technique: Stack-aware template generation. The wizard auto-detects tech stack from the repository URL and fills in conventions accordingly — reducing manual CLAUDE.md authoring from 30 minutes to 2 minutes.

Excerpt 2 — Blueprint with Variables

LynxPrompt supports template variables in blueprints:

# {{PROJECT_NAME}} Development Guide

**Stack**: {{FRONTEND_FRAMEWORK}}, {{BACKEND_FRAMEWORK}}, {{DATABASE}}

## Style Guide
Use {{STYLE_GUIDE}} coding conventions.

## Branch Strategy
Main branch: {{MAIN_BRANCH}}
Feature branches: feature/{{ISSUE_ID}}-{{DESCRIPTION}}

Technique: Parameterized templates. Variables allow one blueprint to serve multiple projects with different stack choices — a parametric configuration management pattern used in infrastructure-as-code tools.

Excerpt 3 — Monorepo Hierarchy Configuration

For monorepos, LynxPrompt supports hierarchical configs:

my-monorepo/
├── CLAUDE.md           ← root blueprint (applies to all packages)
├── packages/
│   ├── frontend/
│   │   └── CLAUDE.md   ← overrides for React/TypeScript frontend
│   ├── api/
│   │   └── CLAUDE.md   ← overrides for Python/FastAPI backend
│   └── shared/
│       └── CLAUDE.md   ← shared utilities conventions

Technique: Configuration inheritance. Root config sets organization-wide rules; package-level configs override for specific contexts. Mirrors how .gitignore and .editorconfig inheritance works.

09

Uniqueness

LynxPrompt — Uniqueness

differs_from_seeds

LynxPrompt is closest to agent-os (both are "write files that agents read" frameworks) and claude-conductor (both are markdown scaffold tools), but LynxPrompt operates at a fundamentally different scale: it is a multi-user, enterprise-ready platform with a full database, authentication stack (SAML/OIDC/LDAP), federated marketplace, and a 18-subcommand CLI — where agent-os and claude-conductor are single-developer directory bundles. Unlike all seeds, LynxPrompt treats AI config files as organizational artifacts requiring version control, team sharing, and compliance management. The closest seed parallel is spec-kit (which also has a CLI and aims for consistency), but spec-kit governs agent behavior in sessions while LynxPrompt governs config file distribution across machines and teams. LynxPrompt has no behavioral enforcement mechanism — it doesn't care what the agent does with the CLAUDE.md, only that teams have a consistent, governed way to create and distribute it.

Positioning

  • Primary differentiator: The only framework in the batch (and likely in the broader catalog) that provides enterprise-grade governance for AI configuration files — SSO, federated marketplace, multi-user team management, version control.
  • Secondary differentiator: Format-agnostic distribution: write once, export to 30+ AI tool formats. This is similar to a "design token" system for AI configs.
  • Target user: Engineering teams at organizations with multiple developers, multiple AI tools, and a need for standardized AI coding assistant behavior across projects.

Observable Failure Modes

  1. Infrastructure overhead: Docker Compose + PostgreSQL is heavy for solo developers; codesight or agent-os serve the same config generation need with zero infrastructure.
  2. GPL-3.0 license: Enterprise adoption may be blocked by GPL-3.0 copyleft; other tools in the batch are MIT or Apache-2.0.
  3. Platform dependency: Teams become dependent on the self-hosted LynxPrompt server for config distribution — if the server goes down, lynxp pull fails.
  4. No behavioral enforcement: LynxPrompt distributes configs but cannot enforce that agents follow them — a Claude Code session can still ignore the distributed CLAUDE.md.
  5. Feature surface complexity: Stripe marketplace, blog, forum, n8n integration, Helm chart, Chocolatey — extensive feature surface for a config distribution tool; may create maintenance burden.
04

Workflow

LynxPrompt — Workflow

Primary Workflow: Config Generation + Distribution

Phase 1: Setup (Server)

# Self-host
docker compose -f docker-compose.selfhost.yml up -d
# → http://localhost:3000

Phase 2: Authentication

lynxp login   # opens browser on configured instance

Phase 3: Blueprint Creation

# Via CLI wizard
lynxp wizard
# Or via web UI (auto-detects stack from GitHub URL)

Phase 4: Distribution

# Pull a blueprint
lynxp pull bp_abc123

# Search for relevant blueprints
lynxp search "next.js typescript testing"

# List available
lynxp list

Phase 5: Sync

# Push local changes back
lynxp push

# Check drift between local and cloud
lynxp diff
lynxp status

Phase-to-Artifact Map

Phase Artifact
wizard CLAUDE.md, AGENTS.md, .cursorrules, etc. (format-specific)
pull Local AI tool config files in correct workspace paths
push Blueprint version in LynxPrompt server
diff Human-readable drift report

Approval Gates

None automated — the platform is a distribution mechanism, not an agent executor. Users explicitly pull/push configs via CLI or web UI.

Team Onboarding Workflow

# Team lead creates blueprint
lynxp wizard → push

# New team member onboards
lynxp login
lynxp pull bp_team-standard
# → AI tool configs installed in project

CI/CD Integration

Via GitHub Action:

- uses: GeiserX/lynxprompt-action@v1
  with:
    blueprint_id: bp_abc123
    api_url: https://lynxprompt.company.com
    api_token: ${{ secrets.LYNXPROMPT_TOKEN }}

Format Conversion

lynxp convert --from claude-md --to cursorrules input.md
lynxp merge bp_1 bp_2 --output merged.md
06

Memory Context

LynxPrompt — Memory & Context

State Storage

LynxPrompt uses PostgreSQL as its primary state store.

Server-Side State

Store Content
PostgreSQL (via Prisma) Blueprints, users, teams, marketplace listings, blueprint versions
File system Generated AI config files (local export)

Client-Side State (CLI)

File Content
~/.config/lynxprompt/ (or lynxp config path) CLI config: API URL, auth token
Local project files Pulled blueprint outputs (CLAUDE.md, AGENTS.md, etc.)

Persistence Scope

global — blueprints live on the LynxPrompt server, accessible from any machine with CLI access.

Cross-Session Handoff

Blueprints are version-controlled on the server. Running lynxp pull bp_abc123 on a new machine restores the exact same config that was created on another machine.

Diff Tracking

lynxp diff and lynxp status track drift between local files and the server-side blueprint version — similar to git status for AI config files.

Context Compaction

Not applicable — LynxPrompt is a distribution platform, not an agent executor. It does not manage AI session context directly.

Blueprint Versioning

Server-side blueprint versions are maintained with the Prisma ORM. The exact version history UI features are not detailed in the README.

Data Export API

/api/export — full JSON data export available for backup or migration.

07

Orchestration

LynxPrompt — Orchestration

Multi-Agent Pattern

None — LynxPrompt is a config management platform, not an agent orchestrator. It does not spawn agents or coordinate AI tasks.

Orchestration Pattern

none

Execution Mode

interactive-loop — the web UI is a persistent server; CLI commands are one-shot invocations against the server API.

Isolation Mechanism

none — generates files to local filesystem; no sandboxing.

Multi-Model Usage

Yes (optional) — the ENABLE_AI flag enables Anthropic API calls for AI-assisted blueprint editing:

AI_MODEL=claude-3-5-haiku-latest  # configurable
ANTHROPIC_API_KEY=sk-...          # user-provided

This is the only AI call in the platform — blueprint generation and distribution are otherwise model-free.

Multi-Platform Routing

No — AI calls go only to Anthropic (when enabled). The platform itself is model-agnostic in what configs it stores and distributes.

MCP Server

lynxprompt-mcp is a separate repository providing an MCP server for browsing and fetching blueprints from AI assistants. This allows Claude Code or other MCP-compatible tools to access blueprints directly.

Cross-Tool Portability

High — the output (AI config files) is consumed by 30+ AI tools. The platform itself is tooling-agnostic; it manages the files that tools read.

Consensus Mechanism

None.

Prompt Chaining

No — blueprint distribution is discrete (pull/push) rather than a chained pipeline.

08

Ui Cli Surface

LynxPrompt — UI & CLI Surface

CLI Binary

  • Binary name: lynxp (alias: lynxprompt)
  • Is thin wrapper: No — full CLI client against the LynxPrompt API
  • Install: npm install -g lynxprompt
  • Subcommands: 18 (wizard, pull, push, search, list, status, diff, convert, merge, hierarchies, link, import, analyze, check, login, logout, whoami, config)
lynxp wizard              # Interactive config generator
lynxp pull bp_abc123      # Pull blueprint to local files
lynxp push                # Push local configs to server
lynxp search "query"      # Search blueprints
lynxp diff                # Diff local vs cloud
lynxp status              # Show drift
lynxp convert ...         # Convert between formats
lynxp merge ...           # Merge blueprints
lynxp login               # Authenticate
lynxp config              # View/edit CLI config
lynxp config set-url ...  # Point to self-hosted instance

Web Dashboard

  • Exists: Yes
  • Type: web-dashboard (Next.js 16)
  • Port: 3000 (default self-hosted)
  • Tech stack: Next.js 16, App Router, PostgreSQL/Prisma, NextAuth.js, Radix UI, React Hook Form, Tailwind CSS
  • Features:
    • Blueprint marketplace (search, browse, favorite)
    • Interactive config wizard
    • Format export (30+ AI tools)
    • Blueprint versioning and diff
    • User/team management
    • Admin panel
    • Analytics (optional Umami)
    • Optional blog, forum, Stripe marketplace

VS Code Extension

  • Exists: Yes — LynxPrompt.lynxprompt on VS Code Marketplace
  • Features: Browse cloud blueprints + local configs in sidebar, pull configs into workspace, diff local vs cloud, push updates

MCP Server

  • Exists: Yes (separate repo: lynxprompt-mcp)
  • Purpose: Browse and manage blueprints from AI assistants via Model Context Protocol

GitHub Action

  • GeiserX/lynxprompt-action@v1 — CI/CD integration

n8n Integration

  • n8n-nodes-lynxprompt — workflow automation for blueprint operations

Kubernetes

  • Helm chart at charts/lynxprompt/ — production-grade Kubernetes deployment

Related frameworks

same archetype · same primary tool · same memory type

claude-mem (thedotmack) ★ 78k

Background worker service captures every tool call as an observation, AI-compresses sessions, and auto-injects relevant past…

pi (badlogic/earendil) ★ 55k

A minimal, hackable, multi-provider terminal coding agent that adapts to your workflows via npm-installable TypeScript Extensions…

Agent Skills (Addy Osmani) ★ 46k

Encodes senior-engineer software development lifecycle as 23 auto-routed skills and 7 slash commands for any AI coding agent.

wshobson/agents Plugin Marketplace ★ 36k

Single Markdown source for 83 domain-specialized plugins that auto-generates idiomatic artifacts for five AI coding harnesses.

TabbyML/Tabby ★ 34k

Self-hosted AI coding assistant server (alternative to GitHub Copilot) with admin dashboard, RAG-based completions, and multi-IDE…

Compound Engineering ★ 17k

Make each unit of engineering work compound into easier future work via brainstorm→plan→execute→review→learn cycles.