convex-agent-skillz MCP — Cas d'usage, Installer & Démo en direct

Pourquoi l'utiliser

Fonctionnalités clés

Deep Convex patterns — queries, mutations, actions, HTTP endpoints
Real-time features: presence, streaming, collaborative editing
Durable ops: workflows, queues, scheduled tasks
Auth patterns — universal, plus Clerk and WorkOS specifics
AI agent patterns: threads, tool calling, RAG, persistent message history

Démo en direct

Aperçu en pratique

convex-agent-skillz-skill.replay ▶ prêt

0/0

Installer

Choisissez votre client

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "convex-agent-skillz-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/PolarCoding85/convex-agent-skillz",
        "~/.claude/skills/convex-agent-skillz"
      ],
      "_inferred": true
    }
  }
}

Ouvrez Claude Desktop → Settings → Developer → Edit Config. Redémarrez après avoir enregistré.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "convex-agent-skillz-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/PolarCoding85/convex-agent-skillz",
        "~/.claude/skills/convex-agent-skillz"
      ],
      "_inferred": true
    }
  }
}

Cursor utilise le même schéma mcpServers que Claude Desktop. La config projet l'emporte sur la globale.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "convex-agent-skillz-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/PolarCoding85/convex-agent-skillz",
        "~/.claude/skills/convex-agent-skillz"
      ],
      "_inferred": true
    }
  }
}

Cliquez sur l'icône MCP Servers dans la barre latérale Cline, puis "Edit Configuration".

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "convex-agent-skillz-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/PolarCoding85/convex-agent-skillz",
        "~/.claude/skills/convex-agent-skillz"
      ],
      "_inferred": true
    }
  }
}

Même format que Claude Desktop. Redémarrez Windsurf pour appliquer.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "convex-agent-skillz-skill",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/PolarCoding85/convex-agent-skillz",
        "~/.claude/skills/convex-agent-skillz"
      ]
    }
  ]
}

Continue utilise un tableau d'objets serveur plutôt qu'une map.

~/.config/zed/settings.json

{
  "context_servers": {
    "convex-agent-skillz-skill": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/PolarCoding85/convex-agent-skillz",
          "~/.claude/skills/convex-agent-skillz"
        ]
      }
    }
  }
}

Ajoutez dans context_servers. Zed recharge à chaud à la sauvegarde.

claude mcp add convex-agent-skillz-skill -- git clone https://github.com/PolarCoding85/convex-agent-skillz ~/.claude/skills/convex-agent-skillz

Une seule ligne. Vérifiez avec claude mcp list. Supprimez avec claude mcp remove.

Cas d'usage

Usages concrets : convex-agent-skillz

Add real-time presence to an existing Convex app

👤 Convex devs adding collab features ⏱ ~45 min intermediate

Quand l'utiliser : You have an app and want Figma-style 'who's here' indicators without running a socket server.

Prérequis

Skill copied into your project's .claude folder — git clone https://github.com/PolarCoding85/convex-agent-skillz and copy .claude/ to your project root

Déroulement

Describe the feature

Using the Convex skillz — add presence to the documents page. Show cursors and usernames.✓ Copié

→ Claude implements with Convex's presence pattern (mutation + reactive query), not a custom WebSocket
Handle disconnection

What about cleanup when users close the tab?✓ Copié

→ Specific stale-presence cleanup strategy aligned with Convex conventions

Résultat : Working presence with proper cleanup.

Pièges

Custom presence pattern that fights Convex's reactivity — Skill steers you toward the idiomatic version; don't override without a reason

Combiner avec : filesystem

Integrate Clerk auth with Convex properly

👤 Devs wiring Clerk into a new or existing Convex app ⏱ ~30 min intermediate

Quand l'utiliser : You picked Clerk and want its Convex integration right the first time.

Déroulement

Wire the provider

Convex skillz — integrate Clerk auth with Convex. Set up the JWT template and the Convex auth config.✓ Copié

→ Claude references the right Clerk JWT template name and auth.config.ts setup
Protect queries

Now gate the documents query so only the owner can read.✓ Copié

→ ctx.auth.getUserIdentity() pattern with proper identity check

Résultat : Auth that passes code review and doesn't leak across tenants.

Build an AI agent with persistent threads on Convex

👤 Devs building chat or agent products ⏱ ~90 min advanced

Quand l'utiliser : You want threaded message history that survives across sessions.

Déroulement

Model threads

Convex skillz — agent threads table + message history + tool calling metadata.✓ Copié

→ Schema that covers streaming, partial messages, and tool-call state
Hook up streaming

Stream the assistant response back via a reactive query.✓ Copié

→ Partial-message pattern leveraging Convex's reactivity

Résultat : A Convex-native agent thread layer ready to scale.

Combinaisons

Associez-le à d'autres MCPs pour un effet X10

convex-agent-skillz-skill + github

Open PRs for feature additions with Convex-conventional diffs

Implement the presence feature, commit and open PR with the Convex skill patterns noted.✓ Copié

convex-agent-skillz-skill + filesystem

Work across schema, functions, and UI in one session

Update convex/schema.ts, add the mutation, and wire up the React hook.✓ Copié

Outils

Ce que ce MCP expose

Outil	Entrées	Quand appeler
queries_mutations_actions	feature intent	Any convex/ work
realtime_patterns	feature (presence / streaming / collab)	Real-time features
auth_setup	provider (Clerk / WorkOS / custom)	Wiring auth
durable_ops	workflow intent	Long-running or retry-heavy logic
ai_agent_threads	agent feature description	Agent products on Convex

Coût et limites

Coût d'exécution

Quota d'API: None from the skill
Tokens par appel: Moderate — patterns load as needed by context
Monétaire: Free — skills are local files. Convex itself has its own pricing.
Astuce: Place skill in project .claude/ rather than global — auto-activation works best scoped to the project.

Sécurité

Permissions, secrets, portée

Stockage des identifiants : No credentials in the skill. Your auth provider secrets live in your own env.

Sortie de données : None from the skill — code stays local

Auth patterns assume you configure JWT templates correctly in Clerk/WorkOS — read their docs in addition.

Dépannage

Erreurs courantes et correctifs

Skill not invoked — Claude writes non-Convex-idiomatic code

Say 'use the Convex skillz' or ensure you're working inside convex/ directory.

Vérifier : ls <project>/.claude/agents/convex-expert.md

SKILL.md frontmatter wrong

Re-copy .claude/ from the repo.

Vérifier : head <project>/.claude/skills/*/SKILL.md

Wrong trigger keywords — auth pattern doesn't use your provider

Name the provider explicitly ('with Clerk' / 'with WorkOS').

Alternatives

convex-agent-skillz vs autres

Alternative	Quand l'utiliser	Compromis
Convex official docs	You want primary source rather than a curated skill	No in-agent auto-activation
Hand-rolled patterns	You're an experienced Convex dev	You'll reinvent patterns; skill just makes Claude use yours consistently

Plus

Ressources

📖 Lire le README officiel sur GitHub

🐙 Voir les issues ouvertes

🔍 Parcourir les 400+ serveurs MCP et Skills