https://promptmarket.sh/mcpConnect any client that supports remote MCP over Streamable HTTP. Use the configuration format for your agent; merge the entry into an existing file instead of replacing your other settings.
Codex
Project config: .codex/config.toml
[mcp_servers.promptmarket]
url = "https://promptmarket.sh/mcp"Codex loads project configuration only after you trust the project. Reload your session after setup. Official setup docs
OpenCode
Project config: opencode.json
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"promptmarket": {
"type": "remote",
"url": "https://promptmarket.sh/mcp",
"enabled": true
}
}
}An existing opencode.jsonc is also supported. Reload OpenCode after setup. Official setup docs
Cursor
Project config: .cursor/mcp.json
{
"mcpServers": {
"promptmarket": {
"url": "https://promptmarket.sh/mcp"
}
}
}Enable PromptMarket in Cursor’s MCP settings after setup. Official setup docs
Claude Code
Project config: .mcp.json
{
"mcpServers": {
"promptmarket": {
"type": "http",
"url": "https://promptmarket.sh/mcp"
}
}
}Start Claude Code in this project and approve the project MCP server when prompted. Official setup docs
GitHub Copilot (VS Code)
Project config: .vscode/mcp.json
{
"servers": {
"promptmarket": {
"type": "http",
"url": "https://promptmarket.sh/mcp"
}
}
}Use MCP: List Servers in VS Code to start and verify PromptMarket. This setup targets Copilot in VS Code. Official setup docs
Other agents
Add a remote HTTP MCP server named promptmarket with the endpoint above, using your client’s settings or config schema. These formats are client-specific; there is no universal MCP config file.
If your agent has no remote MCP support, pass it context from the CLI:
pnpm dlx @promptmarket/cli context "add RAG over our documentation" --project . --format agentThe CLI setup also adds workflow instructions to your agent’s project instruction file. Manual MCP configuration only connects the tools. Context7 is optional; use it or official library documentation to verify current syntax.
Primary tools: get_workflow, build_context, build_plan, search_prompts, get_prompt, search_learn, get_learn_topic, search_guides, get_guide, and recommend_prompt.
build_context assembles the lesson, prompt, guide, and skill for a feature in one call. recommend_prompt returns one match only when the wording is specific.
Skill tools remain available: search_recipes, inspect_recipe, get_recipe, and list_recipe_versions. Browse them on Skills.