Make README agent-agnostic (not just Claude Code)#19
Conversation
The MCP server works with any MCP client, and the skills also ship as
the @valtown/skills npm library. Reframe the README around that:
- Title: "Claude Code plugin for Val Town" -> "Val Town for AI coding agents"
- Lead with the MCP server as an open standard; Claude Code becomes one
of several clients (add Codex CLI, Cursor, VS Code, generic MCP)
- Document the @valtown/skills library usage
- Drop Claude-Code-specific asides in Contributing ("audience-neutral"
-> "agent-neutral")
Modeled on Vercel's MCP docs (supported-clients pattern) and Resend's
concise install/usage style.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
![devin-ai-integration[bot]](https://avatars.githubusercontent.com/in/811515?s=60&v=4){kind=small}
| @@ -1,32 +1,92 @@ | |||
| # Claude Code plugin for Val Town | |||
| # Val Town for AI coding agents | |||
There was a problem hiding this comment.
I suspect there's probably value in keeping the name "Claude Code" in here; maybe also adding in Codex et al? Or maybe listing a few prominently higher up in the README?
There was a problem hiding this comment.
I agree that listing a few higher up is a good idea.
I think keeping plugin in the name is a good idea, probably.
I'd be fine with (Claude Code, Codex, etc) in the name, but maybe just at the top but not in the name is cleaner.
Prior art:
jxnblk
left a comment
jxnblk
left a comment
There was a problem hiding this comment.
General direction SGTM. Left a few minor comments
| cron/intervals, SQLite, email, OAuth, React UI, third-party integrations, | ||
| templates). | ||
| - **The Val Town MCP server** — tools to create, edit, run, and deploy vals, | ||
| query SQLite and blob storage, read logs and traces, and more. It speaks the |
There was a problem hiding this comment.
This whole "It speaks" sentence sounds weird to me; could be a footnote
| external agents alike. | ||
|
|
||
| ## Install | ||
| The MCP server is hosted at `https://api.val.town/v3/mcp` and serves both the |
There was a problem hiding this comment.
This seems like it could be an option at the end of the "Connect" section; something like "Connect directly to the MCP server"
|
|
||
| ## Install | ||
| The MCP server is hosted at `https://api.val.town/v3/mcp` and serves both the | ||
| tools and the skill content (via its `find_val_town_skills` tool), so any agent |
There was a problem hiding this comment.
I'm not sure that the end-user needs to know that we have these skills + a find skills tool. Ideally it should be sort of invisible to the user
2 participants
Reframes the README around the fact that this works with any AI coding agent, not just Claude Code.
The MCP server (
https://api.val.town/v3/mcp) is a standard OAuth-backed remote MCP server that works with any MCP client, and it serves both the tools and the skill content (viafind_val_town_skills). The Claude Code plugin is just a convenience wrapper. The skills also ship as the@valtown/skillsnpm library, which the old README never mentioned.Changes
## Connectsection with per-client setup: Claude Code, Codex CLI, Cursor, VS Code, and a generic "any other MCP client" catch-all@valtown/skillsModeled on Vercel's MCP docs (supported-clients pattern) and Resend's concise install/usage style.
Note
The Codex/Cursor/VS Code snippets follow each tool's documented syntax but haven't been smoke-tested against Val Town's MCP OAuth flow specifically — worth a quick verify before relying on them.
🤖 Generated with Claude Code