Skip to main content
MCP server CX Codex

Warmplane — Claude Code MCP server: proxy and policy layer

A local proxy that keeps MCP connections warm, hands the agent a compact tool catalog, and routes every call through policy checks.

Open on GitHub

What Warmplane does

Warmplane sits between your AI client and a set of upstream MCP servers. Rather than shipping full JSON schemas for every tool on every turn, it exposes a small capability index and lets the agent pull a schema when it actually needs one. SHA-256 ETag caching stops the same payload from being sent twice.

Governance lives in the same process. Config keys allow, deny, requireApproval and redactKeys decide what runs and what gets masked, profiles group servers by task, and multi-tenant RBAC covers shared setups. Retries after a network hiccup are handled with exactly-once deduplication keyed by idk_<sha256>, plus safe, idempotent and unsafe classifications. A supervisor with circuit breakers restarts upstream processes that hang or die.

Three run modes are available: a child process via warmplane mcp-server, a background daemon with warmplane daemon --port 9090 that also serves a REST API and web panel, or a shared instance over HTTP/SSE at http://127.0.0.1:9191/sse. Install through Homebrew (brew tap warmplane/tap) or cargo install warmplane; semantic search needs the extra --features semantic-search build. No Rust toolchain and no Homebrew means no install.

What it helps with

  • cut prompt tokens spent on tool schemas
  • override a misleading tool description without touching upstream code
  • avoid running the same mutation twice after a dropped connection
  • gate risky calls behind human approval
  • strip tokens and passwords out of responses and logs

Example prompts for the agent

  • Add the filesystem server: warmplane server add filesystem --command npx --arg "-y" --arg "@modelcontextprotocol/server-filesystem" --arg "/tmp"
  • Pull in my existing servers with warmplane config import
  • Start the daemon on port 9090 and list what's in the capability catalog

How to install Warmplane

Install

The MCP registry has no launch command for it: see the repository README

Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more

Questions about Warmplane

How do I install Warmplane?

With Homebrew: brew tap warmplane/tap, then brew install warmplane. Or with Cargo: cargo install warmplane, adding --features semantic-search if you want local vector search.

What does Warmplane do?

It proxies MCP traffic locally. Connections to upstream servers stay open, the agent sees a compact catalog instead of full schemas, and each call passes through your policy rules.

How is Warmplane different from connecting MCP servers directly?

Direct connections resend full schemas every turn and apply no checks. Warmplane caches schemas, deduplicates retries, and can require approval before a mutation runs.

Similar MCP servers

All MCP servers for Agent setup: memory, context, orchestration →

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor