LynxPrompt is a self-hosted web application whose core purpose is to solve the problem of inconsistent rule file formats across AI coding tools: letting one set of rules automatically adapt to the file formats required by different AI IDEs and tools. It can generate a total of 30 rule file formats, including AGENTS.md, CLAUDE.md, .cursor/rules/, Copilot instructions, and more, and additionally generates slash commands for six of those formats.
Main capabilities include: 1) A 12-step wizard (Web or terminal) builds rules based on the project, tech stack, commands, code style, and boundary conditions; lynxp analyze can first analyze an existing repository. 2) Blueprints support [[VARIABLE|default]] placeholders, making the same blueprint applicable to multiple projects; each blueprint can be set to private, team-shared, or public, teams have their own dedicated pages, and federated instances can share public blueprints with each other. 3) Hierarchical AGENTS.md is provided for monorepos and can be managed through the dashboard, CLI, and API. 4) The lynxp CLI provides 19 commands (such as wizard, check, analyze, convert, diff, pull, push, search, etc.), and the VS Code extension can be used with any instance.
How to use: You can self-deploy via Docker Compose or a Helm chart, or try the hosted instance lynxprompt.com first. By default, set ADMIN_EMAIL to your own email and open http://localhost:3000，用该邮箱登录，并在 http://localhost:8025（Mailpit，本地邮箱）读取登录链接；该账户即为超级管理员。正式部署时需在 . Set the SMTP_* variables in .env and remove the Mailpit file. The image is amd64 and runs on ARM through emulation. For Helm, the CLI, and other installation channels, see Getting started. Login methods support magic link, passkeys, GitHub or Google OAuth, as well as SAML, OIDC, and LDAP SSO, each requiring only one environment variable.
Use cases: development scenarios where teams want to centrally manage and distribute AI coding rules, need to keep rules consistent across multiple AI coding tools, or need to batch-generate and sync rule files for multiple repositories. In terms of limitations, the AI editing feature is optional, requires your own Anthropic API key, and is disabled by default; all other features can be used without that key. Documentation is at geiserx.github.io/LynxPrompt. The project is licensed under AGPL-3.0-or-later.
