Connect to agent
Connect your agent to the InFlow CLI installed on your computer. MCP lets the agent use InFlow commands as tools. Your agent application runs inflow --mcp to start InFlow's local tool server. Skills provide instructions for using those tools or running the CLI directly. Start with the setup prompt below, or follow the instructions for your agent application.
Fetch https://inflowcli.ai/skill.md and help me connect this InFlow installation to my agent environment.
The SETUP GUIDE walks your agent through connecting to InFlow. The agent needs permission to run commands and edit configuration files on your computer. If it cannot do that, use the instructions below to set up the connection yourself.
Set up your agent application
| Application | Method | Instructions |
|---|---|---|
| Local coding agent | Paste a prompt | Paste the prompt above to load the setup guide at inflowcli.ai/skill.md |
| Install skills | npx skills add inflowpayai/inflow-cli | |
| Claude Code | Plugin | /plugin marketplace add inflowpayai/inflow-cli/plugin install inflow@inflow |
| Codex | Plugin | codex plugin marketplace add inflowpayai/inflow-clicodex plugin add inflow@inflow |
| Cursor | MCP and skills | Add the JSON below to .cursor/mcp.json in your project. Use the skills installation command above and select Cursor. |
Cline Continue Claude Desktop | MCP configuration | Use the configuration and file locations in the MCP configuration section below. |
| Payment instructions | inflow --skill agentic-payments prints instructions you can copy into your agent's chat. On macOS, append | pbcopy to copy them to the clipboard. | |
| OpenClaw | Install skills | npx skills add inflowpayai/inflow-cli — select OpenClaw and the discovery, enrollment, and payment skills. |
| Claude.ai | No local connection | The website cannot start InFlow on your computer. Use an application that supports local MCP servers, such as Claude Desktop. |
| GitHub Copilot | Local MCP server | In VS Code, run MCP: Add Server, choose Command (stdio), and enter inflow --mcp. |
MCP configuration
This configuration tells your agent application how to start InFlow. If you already have an mcpServers section, add the inflow entry without replacing your other servers.
For Continue, save the JSON in .continue/mcpServers/inflow.json in your project. For Cline or Claude Desktop, open the application's MCP configuration editor and add the entry there.
{
"mcpServers": {
"inflow": {
"command": "inflow",
"args": ["--mcp"]
}
}
}If you installed skills with npx skills add inflowpayai/inflow-cli, run it again after updating the CLI to refresh those instructions. inflow --skill prints the payment instructions included with your installed version.
Confirm the connection
Ask the agent to list the tools provided by InFlow, such as balances_list and directory_search. Your application may add a prefix to their names. If the tools are missing, check that the InFlow server is enabled in its MCP settings and look for errors in its MCP log. The application must be able to find InFlow on PATH. If a tool reports a locked vault, run inflow vault unlock in your terminal, then ask the agent to retry. For other errors, follow the message returned by the tool.