Integration Guide — Skills Marketplace

Quick Connect

Add this to your project's .mcp.json to give any AI agent access to Skills Marketplace tools. No API keys, no login, no tokens required.

.mcp.json

        {

          "mcpServers": {

            "skills": {

              "type": "http",

              "url": "https://skills.thinqmesh.com/api/mcp"

            }

          }

        }

What This MCP Endpoint Provides

The Skills Marketplace exposes a standard MCP (Model Context Protocol) endpoint over HTTP. Any AI client that supports MCP can connect and access skill metadata, install commands, documentation, and tool information — all without authentication.

Property	Value
Endpoint	`https://skills.thinqmesh.com/api/mcp`
Protocol	MCP JSON-RPC 2.0 over HTTP
Transport	HTTP POST (JSON-RPC) + GET (health check)
Authentication	None — open access for all agents
CORS	Enabled for all origins
Tools	7 tools: ccs-info, ccs-commands, ccs-command-help, ccs-install, ccs-models, ccs-context-files, ccs-docs

Available MCP Tools

These tools provide skill information and setup guidance. The actual codebase operations happen locally via installed skills.

Tool	Purpose	Input
`ccs-info`	Skill overview, features, install commands	None
`ccs-commands`	List all 22 slash commands with model assignments	`category`: all, context, workflow, testing, git, operations
`ccs-command-help`	Detailed usage for a specific command	`command`: init, plan, build, fix, branch, pr, etc.
`ccs-install`	Install command for any OS / package manager	`method`: npx, bunx, pnpm, global, curl, powershell, manual
`ccs-models`	Model strategy — which model handles which task	None
`ccs-context-files`	Description of all .ccs/ context files	None

Client Setup

Claude Code

Full native skill support. Install locally for all 22 slash commands.

Claude Code

        # Install skills locally (recommended)

        $ npx codebase-context-skill init

        # Or connect via MCP for tool access without local install

        $ claude mcp add --transport http skills https://skills.thinqmesh.com/api/mcp

Codex CLI

Use CCS context files as reference and connect via MCP for tool metadata.

Codex CLI

        # Install CCS, run /ccs-init, then reference .ccs/ files

        $ npx codebase-context-skill init

        $ codex "Read .ccs/file-index.md first, then fix the login bug"

ChatGPT / Cursor / Any MCP Client

Any tool that speaks MCP JSON-RPC over HTTP can connect. Point it at the endpoint — no auth needed.

Generic MCP

        # Health check

        $ curl https://skills.thinqmesh.com/api/mcp

        # List tools

        $ curl -X POST https://skills.thinqmesh.com/api/mcp \

          -H "Content-Type: application/json" \

          -d '{"jsonrpc":"2.0","method":"tools/list","id":1}'

        # Call a tool

        $ curl -X POST https://skills.thinqmesh.com/api/mcp \

          -H "Content-Type: application/json" \

          -d '{"jsonrpc":"2.0","method":"tools/call","params":{"name":"ccs-info","arguments":{}},"id":2}'

After Connecting

Call ccs-info to get the install command for your OS
Install locally with npx codebase-context-skill init
Run /ccs-init to generate the full context index in .ccs/
Every future task: read .ccs/file-index.md first, then follow the Agent Guide workflow

After one setup, the AI agent has everything locally. No more web fetches. No redundant exploration. Every query is context-aware from the first token.