Skip to main content

Overview

The ChainStream skill package (chainstream-io/skills) is distributed through the Vercel Skills CLI, which supports 40+ AI agents. A single command installs and configures the ChainStream MCP tools for your agent of choice.

Compatibility Table

The table below lists the most commonly used agents with their install commands and configuration file locations.

Agent-Specific Details

OpenClaw

The Skills CLI writes the configuration to the OpenClaw skills directory:
  • Project-level: ./skills/chainstream.yaml
  • User-level: ~/.openclaw/skills/chainstream.yaml
Generated configuration:
Set the environment variable before running your agent:

Claude Code

The Skills CLI adds the MCP server entry to .claude.json in your project root:
See also: Stdio Transport for manual Claude Code setup.

Cursor

The Skills CLI configures the MCP server in Cursor’s settings. The configuration appears under Settings > Features > MCP Servers: For project-level config, the CLI writes to .cursor/mcp.json:

Cline

The Skills CLI adds the server to Cline’s MCP configuration:

GitHub Copilot

The Skills CLI writes configuration to .github/copilot/mcp.json in your repository:

Codex

The Skills CLI adds the MCP server to your Codex agent configuration:

Auto-Detection

When you run npx skills add chainstream-io/skills without the -a flag, the CLI auto-detects your environment by checking for:
  1. Agent-specific config files in the current directory
  2. Agent CLI tools installed globally
  3. IDE-specific directories and settings
If multiple agents are detected, the CLI prompts you to choose.

Updating

To update the ChainStream skill to the latest version:

Removing

To remove the ChainStream skill:
This removes the configuration entry from your agent’s config file but does not delete your API key.

Next Steps

Tools Reference

Explore all 15 MCP tools available after installation.

Security

Security considerations for the ChainStream integration.