The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
Mcpshim
Turn remote MCP servers into local command workflows.
Install
go install github.com/mcpshim/mcpshim/cmd/mcpshimd@latestInstall command taken from the project's README or registry entry. Always review code before running it.
README excerpt
Read full README on GitHub →MCPShim
<strong>Use any MCP server or HTTP API as a standard CLI command.</strong>
A lightweight daemon + CLI that turns remote MCP tools and configured HTTP endpoints into native shell commands your agent or script can call directly.
Website · Repository · Quick Start · Core Commands
The Problem
Remote MCP servers and HTTP APIs are powerful, but each service has its own auth flow, transport expectations, and invocation patterns. Wiring all of that directly into every script or agent loop creates brittle command workflows.
For LLM agents, there is also context pressure: dumping raw MCP schemas for every connected server can consume prompt budget before useful work begins.
The Solution
mcpshimd centralizes MCP registration and OAuth alongside configured HTTP
service bindings, discovery, call execution, and history behind one local
socket.
mcpshim exposes every remote MCP tool and configured HTTP operation as a
standard CLI command. Flags map to tool parameters and output comes back as
structured JSON. No SDKs or libraries are required by the caller.
graph TD
Agent["Your AI Agent / Script"]
Agent -->|call| CLI["mcpshim CLI"]
Agent -->|JSON request| Socket["Unix Socket"]
CLI --> Socket
Socket --> Daemon["mcpshimd"]
Daemon --> MCP1["MCP Server: Notion"]
Daemon --> MCP2["MCP Server: GitHub"]
Daemon --> MCP3["MCP Server: Linear"]
Daemon --> HTTP["Configured HTTP API"]
Daemon --> MCPN["..."]
Why MCPShim
| Without MCPShim | With MCPShim | |
|---|---|---|
| Tool integration | Custom per-service wiring | One daemon + one CLI |
| Auth handling | Per-script OAuth/header logic | Centralized in mcpshimd |
| Tool invocation | Provider-specific conventions | mcpshim call --server --tool ... |
| HTTP API binding | Hand-written client code | Typed or constrained raw tools |
| Agent context budget | Large MCP schemas in prompt | Alias-based local command workflows |
| Operational history | Ad-hoc logging | Built-in call history in SQLite |
Architecture
| Component | Role |
|---|---|
mcpshimd |
Local daemon for MCP/HTTP registry, discovery, auth, calls, and IPC |
mcpshim |
CLI client for config, discovery, tool calls, history, and script |
All client calls go through a Unix socket and JSON request/response protocol.
Quick Start
1. Install from source
go install github.com/mcpshim/mcpshim/cmd/mcpshimd@latest
go install github.com/mcpshim/mcpshim/cmd/mcpshim@latest
2. Configure
mkdir -p ~/.config/mcpshim
cat > ~/.config/mcpshim/config.yaml <<'YAML'
servers:
- name: notion
alias: notion
transport: http
url: https://mcp.notion.com/mcp
YAML
3. Start daemon and inspect
mcpshimd
mcpshim status
mcpshim servers
mcpshim tools
Path Defaults
Similar plugs
Reverse engineer anything with agents, from app behavior down to native binaries.
Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype…
User-friendly AI Interface (Supports Ollama, OpenAI API, ...)
A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Grok Build & Hermes Agent. Only official website: ccswitch.io