Best for
- Developers and AI agents who want to interact with MCP servers via CLI
- Users who need to script MCP operations or integrate into CI/CD
- Those who need to safely expose MCP credentials in sandboxed environments
M
66FMRS · C
mcpc is a feature-rich MCP CLI client with comprehensive MCP support, ideal for developers and AI agents. Its security features (like OAuth and proxy) and JSON output make it excellent for scripting and automation. However, x402 and skills support are still experimental, and it lacks Roots and Sampling support. Overall, it is a powerful and flexible tool, but users should be aware of security risks and the instability of experimental features.
mcpc is a command-line client for the Model Context Protocol (MCP) that maps every MCP operation to an intuitive shell command. It supports persistent sessions, stdio and Streamable HTTP transports, OAuth 2.1 authentication (with CIMD and DCR), tasks, JSON output for code mode, proxy for AI sandboxes, and experimental x402 payments. It is designed to give AI agents full MCP protocol access through a single Bash() tool call, avoiding the need to wire up dozens of MCP functions.
Tools
connect
Connect to an MCP server and start a new named session
close
Close a session
restart
Restart a session (losing all state)
login
Log in to a server and save an OAuth profile
logout
Delete an OAuth profile for a server
clean
Clean up mcpc data (sessions, profiles, logs, all)
grep
Search tools and instructions across all active sessions
x402
Configure an x402 payment wallet (EXPERIMENTAL)
help
Show help for a command
tools-list
List all server tools
tools-get
Get tool details and schema
tools-call
Call a tool with arguments
tasks-list
List active tasks
tasks-get
Check task status
tasks-result
Get task result (blocks until terminal state)
tasks-cancel
Cancel a running task
prompts-list
List available prompts
prompts-get
Retrieve a prompt template
resources-list
List available resources
resources-read
Read a resource content
resources-subscribe
Subscribe to resource updates and sync to a file
resources-unsubscribe
Unsubscribe from resource updates
resources-templates-list
List resource templates
skills-list
List skills exposed by the server (experimental)
skills-get
Read a skill's SKILL.md
logging-set-level
Set server logging level (deprecated)
ping
Send a ping to check connectivity
server-discover
Send server/discover request (2026-07-28 protocol)
logs
View session bridge logs
Setup
Install with Homebrew (macOS/Linux): brew install apify/tap/mcpc; or via npm globally: npm install -g @apify/mcpc (requires Node.js or Bun).
Fit and risk
Not for
- Users seeking a graphical interface
- Users needing built-in LLM calling (mcpc does not use LLMs itself)
- Users requiring Roots or Sampling support (not supported)
Required permissions
- Access local file system for configuration and session files
- Network access to connect to remote MCP servers
- OS keychain for storing credentials
- Potential proxy access in AI sandbox scenarios
Risks and side effects
- MCP servers may execute arbitrary commands or access local resources; ensure to use trusted servers
- Proxy feature may expose local services; configure binding addresses carefully
- x402 payments involve cryptocurrency transactions and financial risk
- OAuth credentials stored in keychain; protect system security
Troubleshooting
- Check session status: run mcpc or mcpc --json to see active sessions and their states
- If a session is unauthorized, re-run mcpc login <server> and restart the session
- If a stdio server fails to start, inspect ~/.mcpc/logs/bridge-<session>.log for errors
- Ensure environment variables like HTTPS_PROXY or SSL_CERT_FILE are set correctly (stdio servers inherit only a minimal env)
Use cases
Manually inspect and debug MCP servers
Automate repeatable MCP workflows in shell scripts
Provide AI agents with full MCP protocol access
Search across sessions to enable progressive tool discovery
Supported clients
Claude DesktopPartial support
Claude CodePartial support